Method and system for detecting an electrical automation device

Abstract

The application relates to the technical field of electrical automation, and particularly discloses a detection method and system for an electrical automation device, wherein the method comprises the following steps: collecting electrical signals of a mechanical arm, and simultaneously acquiring corresponding mechanical arm control information; identifying an electrical signal segment of a repeated motion section of the corresponding mechanical arm; extracting a nonlinear characteristic parameter; comparing the nonlinear characteristic parameter with pre-determined parameter reference data to obtain a nonlinear characteristic parameter variation; and detecting a current health state of the electrical automation device according to the nonlinear characteristic parameter variation and a pre-set health threshold. The method can sensitively capture nonlinear characteristic changes caused by early wear through a nonlinear analysis method, can more early and accurately identify early wear characteristics of mechanical transmission components online, and thus can realize early warning and evaluation of the health state of the electrical automation device, and can avoid that equipment maintenance lags behind the actual wear process.

Description

Technical Field

[0001] The present application relates to the field of electrical automation technology, and in particular to a detection method and system for electrical automation equipment. Background Art

[0002] In the field of electrical automation equipment, multi-axis robotic arms achieve high-precision motion control through servo motors driving mechanical transmission chains. Their repetitive operations require long-term stability and reliability of transmission components. With the advancement of industrial automation, the complexity of robotic arms' motion trajectories and the dynamic range of their loads continue to expand. Drive systems must adapt to more sophisticated control requirements, and changes in the physical properties of transmission components have become a significant factor affecting equipment performance.

[0003] However, existing detection methods struggle to effectively identify early signs of wear in mechanical transmission components during the continuous operation of electrical automation equipment. Because the electrical signals output by servo drives are subject to multiple interferences from the diverse robotic arm motion tasks, load variations, and environmental factors, traditional linear analysis methods are unable to isolate nonlinear changes caused by subtle wear from the dynamic signals. Conventional monitoring methods rely on periodic maintenance or post-failure repairs, resulting in equipment maintenance lagging behind the actual wear process.

[0004] There is currently no effective technical solution to the above problems. Summary of the Invention

[0005] The purpose of this application is to provide a detection method and system for electrical automation equipment, so as to use nonlinear analysis methods to accurately identify the early wear characteristics of mechanical transmission components online, realize early warning and assessment of the health status of electrical automation equipment, and avoid equipment maintenance lagging behind the actual wear process.

[0006] In a first aspect, the present application provides a detection method for electrical automation equipment, which is applied to electrical automation equipment having a robotic arm, and the method comprises the following steps:

[0007] S1. Collect the electrical signal of the robot arm and obtain the corresponding robot arm control information;

[0008] S2. identifying, from the electrical signal, an electrical signal segment corresponding to a repetitive motion segment of the robotic arm based on the robotic arm control information;

[0009] S3, performing nonlinear analysis on the electrical signal segment to extract nonlinear characteristic parameters, wherein the nonlinear characteristic parameters represent the nonlinear characteristics of the mechanical transmission component;

[0010] S4, comparing the nonlinear characteristic parameter with predetermined parameter reference data to obtain a variation of the nonlinear characteristic parameter;

[0011] S5. Detect the current health status of the electrical automation equipment according to the variation of the nonlinear characteristic parameter and a preset health threshold.

[0012] The present invention's detection method for electrical automation equipment, based on robotic arm control information, effectively separates signal segments related to the repetitive motion of mechanical transmission components from dynamic electrical signals subject to interference from various factors. It also utilizes nonlinear analysis to sensitively capture nonlinear characteristic changes caused by early wear. Compared to traditional linear analysis methods, this method can more accurately and earlier identify early wear characteristics of mechanical transmission components online, thereby enabling early warning and assessment of the health status of electrical automation equipment, preventing equipment maintenance from lagging behind actual wear progression.

[0013] In the detection method for electrical automation equipment, the electrical signal includes a current signal and / or a voltage signal of a servo motor of a robotic arm.

[0014] The solution of the present application limits the collected electrical signals to the current signals and / or voltage signals of the servo motor of the robotic arm, so that the signals collected in step S1 can more directly reflect the operating status of the servo motor and the load conditions of the mechanical transmission components.

[0015] In the detection method for electrical automation equipment, the robotic arm control information includes one or more of robotic arm control instructions, joint encoder signals, and motor torque instructions.

[0016] The detection method of the electrical automation equipment, wherein step S3 comprises:

[0017] S31, performing short-time Fourier transform on the electrical signal segment to obtain an electrical spectrum diagram;

[0018] S32. Divide the frequency range into multiple frequency bands based on the vibration characteristics of the robotic arm, and calculate the energy value in each frequency band according to the electrical spectrum diagram to form an energy feature vector;

[0019] S33. Using a pre-trained sparse autoencoder, the energy feature vector is subjected to dimensionality reduction processing to perform nonlinear analysis, thereby obtaining nonlinear feature parameters after dimensionality reduction.

[0020] The detection method of the electrical automation equipment, wherein step S32 comprises:

[0021] S321: Obtain a frequency range relationship table representing the vibration characteristics of the robotic arm, where the frequency range relationship table is a pre-constructed relationship table between the fundamental frequency vibration range and the robotic arm axis;

[0022] S322, dividing the frequency range in the electrical spectrum graph into a plurality of frequency bands according to the frequency range relationship table;

[0023] S323. For each frequency band, calculate the sum of the energies of all frequency components in the frequency band to obtain the energy value corresponding to each frequency band;

[0024] S324. Arrange the energy values ​​of all frequency bands in order of the frequency bands to form an energy feature vector.

[0025] In the detection method for electrical automation equipment, the process of constructing the frequency range relationship table includes:

[0026] A1. Collect historical operating data of the robotic arm under different working conditions, including the types of the robotic arm axes and corresponding electrical signal data;

[0027] A2. For each type of robot arm axis, analyze the electrical signal data collected in step A1, determine the fundamental frequency vibration range of different robot arm axes under different working conditions, establish a corresponding relationship between the fundamental frequency vibration range and the robot arm axis, and form a frequency range relationship table.

[0028] In the detection method for electrical automation equipment, in step S33, the training acquisition process of the sparse autoencoder includes:

[0029] B1. Construct a sparse autoencoder consisting of an input layer, a hidden layer, and an output layer. The number of neurons in the input layer is the same as the dimension of the energy feature vector, the number of neurons in the output layer is the same as the input layer, and the number of neurons in the hidden layer is smaller than the input layer.

[0030] B2. Initialize the weights and bias parameters of the sparse autoencoder and set the coefficient of the L1 regularization term;

[0031] B3. Input the training energy feature vector into the sparse autoencoder, calculate the reconstructed energy feature vector of the output layer, calculate the reconstruction error between the reconstructed energy feature vector and the input energy feature vector, and calculate the value of the L1 regularization term;

[0032] B4. Taking a weighted sum of the reconstruction error and the value of the L1 regularization term to obtain a total loss function value;

[0033] B5. Use the gradient descent method to optimize the weight and bias parameters of the sparse autoencoder to minimize the total loss function value;

[0034] B6. Repeat steps B3-B5 until the total loss function value converges or the maximum number of iterations is reached to obtain a trained sparse autoencoder.

[0035] The detection method for electrical automation equipment, wherein step S2 comprises:

[0036] S21, establish a mapping relationship of timestamps of the mechanical arm control information and the electrical signal;

[0037] S22, analyze the mechanical arm control information to obtain a control instruction subsequence of a repeated motion segment, and determine a start and end timestamp thereof;

[0038] S23, according to the mapping relationship, extract an electrical signal segment corresponding to the start and end timestamps of the control instruction subsequence of the repeated motion segment as an electrical signal segment of the repeated motion segment.

[0039] The detection method of the electrical automation equipment, wherein step S4 comprises:

[0040] S41, obtain pre-determined parameter reference data, the parameter reference data comprising reference values of a plurality of nonlinear characteristic parameters and corresponding weight coefficients;

[0041] S42, compare the nonlinear characteristic parameters with the reference values in the parameter reference data one by one, and calculate a difference value between each nonlinear characteristic parameter and the corresponding reference value;

[0042] S43, according to the weight coefficients, perform a weighted processing on the difference value obtained in step S42 to obtain a weighted change amount of each nonlinear characteristic parameter;

[0043] S44, accumulate the weighted change amounts of all nonlinear characteristic parameters to obtain a nonlinear characteristic parameter change amount.

[0044] In a second aspect, the application further provides a detection system of an electrical automation equipment, applied in an electrical automation equipment with a mechanical arm, the system comprising:

[0045] A collection module for collecting electrical signals of the mechanical arm and simultaneously obtaining corresponding mechanical arm control information;

[0046] A slicing module for identifying an electrical signal segment corresponding to a repeated motion segment of the mechanical arm from the electrical signals based on the mechanical arm control information;

[0047] A feature extraction module for performing nonlinear analysis on the electrical signal segment to extract nonlinear characteristic parameters, the nonlinear characteristic parameters representing nonlinear characteristics of the mechanical transmission components;

[0048] A comparison module for comparing the nonlinear characteristic parameters with pre-determined parameter reference data to obtain a nonlinear characteristic parameter change amount;

[0049] A detection module for detecting a current health state of the electrical automation equipment according to the nonlinear characteristic parameter change amount and a pre-set health threshold.

[0050] The detection system for electrical automation equipment of the present application can identify the early wear characteristics of mechanical transmission components online earlier and more accurately, thereby realizing early warning and assessment of the health status of electrical automation equipment, and avoiding equipment maintenance lagging behind the actual wear process.

[0051] As can be seen from the above, this application provides a detection method and system for electrical automation equipment. Based on robotic arm control information, the method can effectively separate signal segments related to the repetitive motion of mechanical transmission components from dynamic electrical signals that are interfered with by various factors. Furthermore, the method utilizes nonlinear analysis methods to sensitively capture nonlinear characteristic changes caused by early wear. Compared with traditional linear analysis methods, this method can more accurately and earlier identify early wear characteristics of mechanical transmission components online, thereby enabling early warning and assessment of the health status of electrical automation equipment, preventing equipment maintenance from lagging behind the actual wear process. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1 Flowchart of a detection method and system for electrical automation equipment provided in some embodiments of the present application.

[0053] Figure 2 This is a schematic diagram of the structure of the detection method and system for electrical automation equipment provided in some embodiments of the present application.

[0054] Figure 3 A schematic diagram of the structure of a detection method and system for electrical automation equipment provided in some embodiments of the present application.

[0055] Figure 4 This is a schematic diagram of the structure of the detection method and system for electrical automation equipment provided in some embodiments of the present application.

[0056] Figure numerals: 201, acquisition module; 202, slicing module; 203, feature extraction module; 204, comparison module; 205, detection module; 206, prompt module. DETAILED DESCRIPTION

[0057] The technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all of the embodiments. The components of the embodiments of the present application generally described and shown in the drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the application for protection, but merely represents the selected embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without making creative work fall within the scope of protection of the present application.

[0058] It should be noted that similar reference numerals and letters refer to like items throughout the several views, and once an item is defined in one view, it should not require further defining and explaining in subsequent views. Also, in the description of the present application, the terms "first", "second", etc. are used only to differentiate descriptions, and cannot be understood as indicating or implying relative importance.

[0059] In a first aspect, referring to Figure 1 and Figure 2 Some embodiments of the present application provide a detection method of an electrical automation device, applied in an electrical automation device with a mechanical arm, the method comprising the following steps:

[0060] S1, collecting electrical signals of the mechanical arm, and simultaneously acquiring corresponding mechanical arm control information;

[0061] S2, identifying an electrical signal segment of a corresponding mechanical arm repetitive motion segment from the electrical signals based on the mechanical arm control information;

[0062] S3, performing nonlinear analysis on the electrical signal segment to extract a nonlinear characteristic parameter, the nonlinear characteristic parameter representing nonlinear characteristics of a mechanical transmission component;

[0063] S4, comparing the nonlinear characteristic parameter with pre-determined parameter reference data to obtain a nonlinear characteristic parameter variation;

[0064] S5, detecting a current health state of the electrical automation device according to the nonlinear characteristic parameter variation and a pre-set health threshold.

[0065] Specifically, electrical signal refers to the data stream that reflects the state of the robot arm servo drive system, which is mainly used to provide basic data for analyzing the operating state of the robot arm; robot arm control information refers to the data used to control the movement of the robot arm, which is mainly used to provide a motion state reference synchronized with the electrical signal in order to identify specific motion segments; robot arm repetitive motion segment refers to the motion sequence that is periodically repeated when the robot arm performs a specific task, which is mainly used to focus on a stable and repeatable operating state for the convenience of comparative analysis; electrical signal segment refers to the subset of electrical signal data corresponding to the repetitive motion segment of the robot arm, which is mainly used to extract the part related to the specific motion state from the continuous electrical signal and exclude the interference of non-repetitive motion or static state; nonlinear analysis refers to a method of processing signals that can reveal the nonlinear characteristics in the signal, which is mainly used to extract the nonlinear characteristics of the mechanical transmission components from the electrical signal segment. Linear characteristic parameters; nonlinear characteristic parameters refer to numerical values ​​or vectors extracted from electrical signal segments through nonlinear analysis, representing the nonlinear characteristics of mechanical transmission components. Their primary purpose is to quantify the current state of mechanical transmission components, particularly nonlinear changes caused by early wear. Parameter baseline data refers to predetermined nonlinear characteristic parameter values ​​that represent the normal health of the robotic arm. It primarily provides a reference standard for comparing the currently extracted nonlinear characteristic parameters and quantifying their degree of variation. Nonlinear characteristic parameter variation refers to the difference between the current nonlinear characteristic parameters and the parameter baseline data. It primarily quantifies the degree of deviation from the nonlinear characteristics of the mechanical transmission component. Preset health thresholds are threshold values ​​used to determine the health status of electrical automation equipment. They primarily provide a criterion for assessing whether the equipment is normal or abnormal based on the variation in nonlinear characteristic parameters. Preset health thresholds can be set based on historical data or expert experience. When assessing health status, if the variation is less than the threshold, the equipment is considered normal; if it is greater than the threshold, the equipment is considered abnormal or requires attention.

[0066] More specifically, the scheme of the present application provides basic data and motion state reference for subsequent analysis by collecting the electrical signals of the robot arm and the corresponding robot arm control information. Based on the robot arm control information, the method of the present application can accurately identify the electrical signal segments corresponding to the repetitive motion segments of the robot arm, because repetitive motion segments usually represent the core working mode of the robot arm, and their electrical signals can more stably reflect the characteristics of the transmission components, and the control information provides the basis for positioning these specific motion segments. Nonlinear analysis of the extracted electrical signal segments can extract nonlinear feature parameters from the signals. Nonlinear analysis is selected because early wear or performance degradation of mechanical transmission components often first manifests in changes in their nonlinear characteristics, and traditional linear analysis methods are difficult to capture these subtle nonlinear changes. The extracted nonlinear feature parameters characterize the current nonlinear characteristics of the mechanical transmission components. Comparing these nonlinear feature parameters with the pre-determined parameter reference data, the nonlinear feature parameter change is calculated, which quantifies the deviation of the current nonlinear characteristics of the mechanical transmission components from the normal state. Finally, according to the calculated nonlinear feature parameter change and the pre-set health threshold, the method can detect the current health status of the electrical automation equipment. When the change exceeds the threshold, it indicates that the nonlinear characteristics of the mechanical transmission components have changed significantly, which may indicate early wear or potential failure, thereby achieving early warning and assessment of the health status of the equipment. The whole process forms a closed-loop detection process from data collection, specific segment extraction, nonlinear feature extraction, state quantification to final health assessment.

[0067] Through the above technical solutions, the detection method of the electrical automation equipment of the present application can effectively separate the signal segments related to the repetitive motion of the mechanical transmission components from the dynamic electrical signals disturbed by various factors based on the robot arm control information, and sensitively capture the nonlinear feature changes caused by early wear using nonlinear analysis methods. Compared with traditional linear analysis methods, this method can more accurately and earlier identify the early wear characteristics of the mechanical transmission components, thereby achieving early warning and assessment of the health status of the electrical automation equipment, and avoiding equipment maintenance lagging behind the actual wear process.

[0068] In some preferred embodiments, the electrical signals include servo motor current signals and / or voltage signals of the robot arm.

[0069] Specifically, the electrical signal refers to the electrical quantity signal generated by the robotic arm during operation, which can be collected by current sensors, voltage sensors, power meters and other equipment; the servo motor current signal and / or voltage signal of the robotic arm refers to the current signal and / or voltage signal collected directly from the servo motor that drives the joint movement of the robotic arm. The current signal reflects the motor load and torque output, and the voltage signal reflects the motor drive voltage. It can be obtained by using Hall effect current sensors, shunts, voltage divider resistor networks and the like.

[0070] More specifically, the solution of the present application limits the collected electrical signals to the current signals and / or voltage signals of the servo motor of the robotic arm, so that the signals collected in step S1 can more directly reflect the operating status of the servo motor and the load conditions of the mechanical transmission components.

[0071] In some preferred embodiments, the manipulator control information includes one or more of a manipulator control instruction, a joint encoder signal, and a motor torque instruction.

[0072] Specifically, robot control commands are instructions sent to the robot controller or driver. These can be trajectory commands, velocity commands, or position commands, and are used to plan and control the robot's motion. Joint encoder signals are signals output by the joint encoders, used to measure the real-time angles or positions of each joint in the robot. Motor torque commands are instructions sent to the motor driver to control the output torque of the drive motor. This control information provides a more comprehensive and accurate description of the robot's motion state, laying the foundation for the subsequent identification of repetitive motion segments.

[0073] More specifically, the robot arm control instructions directly reflect the robot arm's motion planning. By analyzing the control instruction sequence, it can be determined whether the robot arm is performing repetitive motion. The joint encoder signal provides real-time position information of each joint of the robot arm. By monitoring the changes in joint position, repetitive motion trajectories can be identified. The motor torque instruction reflects the torque required to drive the motor. The change in torque is closely related to the motion state of the robot arm. By analyzing the motor torque instruction, it can assist in identifying repetitive motion segments. This refinement of the control information enables the step of identifying the electrical signal segment corresponding to the repetitive motion segment of the robot arm from the electrical signal based on the robot arm control information to be executed more accurately, thereby improving the accuracy and reliability of the health status detection of the entire electrical automation equipment.

[0074] In some preferred embodiments, the method further comprises the steps of:

[0075] S6. When the current health status is abnormal, generate an abnormal status prompt.

[0076] Specifically, generating an abnormal state prompt refers to generating one or more signals or information by the system when the detection result shows that the current health state of the electrical automation equipment deviates from the normal range to an abnormal degree, which is used to alarm or notify the operator or maintenance personnel, which can be realized in the form of displaying abnormal information on the device operation interface, flashing or changing color through indicator light, issuing a sound alarm, sending a short message or email to the preset contact, or sending an abnormal event report to the remote monitoring system through the network, etc., the purpose of which is to timely inform the relevant personnel that the equipment has an abnormal situation, so that the maintenance personnel can quickly know the potential or existing abnormal situation of the equipment, so that immediate inspection, diagnosis or maintenance intervention measures can be arranged. This timely information feedback mechanism effectively avoids maintenance delay due to information lag, significantly reduces the risk of further deterioration of equipment failure, reduces unplanned downtime, and improves the operation reliability and overall production efficiency of the equipment.

[0077] In some preferred embodiments, step S3 comprises:

[0078] S31, performing short-time Fourier transform on the electrical signal segment to obtain an electrical spectrum diagram;

[0079] S32, dividing the frequency range into multiple frequency bands based on the vibration characteristics of the mechanical arm, calculating the energy value in each frequency band according to the electrical spectrum diagram, and constructing an energy feature vector;

[0080] S33, using a pre-trained sparse autoencoder to perform dimensionality reduction processing on the energy feature vector for nonlinear analysis to obtain nonlinear feature parameters after dimensionality reduction.

[0081] Specifically, the electrical spectrum diagram refers to an image representing the time-frequency domain by converting the time-domain electrical signal segment through short-time Fourier transform, which can be represented in the form of a two-dimensional matrix or an image. The horizontal axis usually represents time, the vertical axis represents frequency, and the color or brightness represents the signal intensity or energy at the corresponding time and frequency point. The purpose is to reveal the frequency components and energy distribution of the signal over time.

[0082] More specifically, dividing the frequency range into multiple frequency bands based on the vibration characteristics of the mechanical arm refers to determining frequency intervals with physical meaning according to specific vibration modes, resonance frequencies or harmonic frequencies generated by the mechanical arm during operation or work, etc. These characteristics can be determined by modal analysis, vibration testing or analysis of historical operation data, etc. to guide the frequency band division, the purpose of which is to focus the complex spectrum information on specific frequency intervals that are more relevant to the health state of the mechanical transmission components.

[0083] More specifically, the energy value refers to the sum of the energy of the signal within each divided frequency band. It can be calculated by integrating or summing the squares of the amplitudes of all frequency components in the frequency band. Its purpose is to quantify the vibration intensity or signal activity of each specific frequency band.

[0084] More specifically, an energy eigenvector refers to a numerical vector formed by arranging the energy values ​​of multiple frequency bands in a specific order. It can be represented by a one-dimensional array or list. Its purpose is to structure the frequency band energy information as input for subsequent nonlinear analysis.

[0085] More specifically, a pre-trained sparse autoencoder refers to a neural network model whose internal parameters (weights and biases) have been adjusted through training data.

[0086] More specifically, dimensionality reduction refers to the process of mapping a high-dimensional energy feature vector to a low-dimensional space through a sparse autoencoder. This can be achieved by inputting the energy feature vector into the hidden layer of a trained sparse autoencoder and extracting the hidden layer output as the dimensionality reduction result. Its purpose is to remove redundancy and noise in the data and extract more essential and discriminative nonlinear features.

[0087] More specifically, the solution of the present application converts the time domain signal into a time-frequency domain representation by first performing a short-time Fourier transform on the electrical signal segment, which makes it possible to observe the changes in the frequency components of the signal over time, laying the foundation for subsequent frequency domain analysis. It is precisely because of the acquisition of the electrical spectrum that the frequency range can be divided into multiple physically meaningful frequency bands based on the vibration characteristics of the robotic arm, and the energy value in each frequency band is calculated, thereby refining the complex spectrum information into an energy feature vector that reflects the activity of a specific mechanical component or vibration mode. It is precisely because of the construction of the energy feature vector that the pre-trained sparse autoencoder can be used to perform nonlinear dimensionality reduction on the vector, thereby extracting low-dimensional, more representative nonlinear feature parameters from high-dimensional, potentially noisy and redundant energy features.

[0088] More specifically, this technical approach, which combines time-frequency analysis, domain-knowledge-guided feature engineering, and nonlinear dimensionality reduction, can more effectively capture subtle nonlinear changes associated with early wear or performance degradation of mechanical transmission components from complex electrical signals. This overcomes the limitations of traditional linear or simple nonlinear methods in processing dynamic, nonlinear system signals. Applying these refined nonlinear feature parameters to health status assessment based on repetitive motion segment signals can significantly improve detection accuracy and sensitivity.

[0089] More specifically, in an embodiment of the present application, the above-mentioned sparse autoencoder is preferably trained with the goal of minimizing the reconstruction error and satisfying the sparsity constraint, wherein the sparsity constraint is implemented by adding an L1 regularization term to the loss function to increase the sensitivity of the feature to wear of mechanical transmission components and suppress the influence of electromagnetic interference.

[0090] In some preferred embodiments, step S32 includes:

[0091] S321. Obtain a frequency range relationship table representing vibration characteristics of the robotic arm, where the frequency range relationship table is a pre-built relationship table between the fundamental frequency vibration range and the robotic arm axis;

[0092] S322, dividing the frequency range in the electrical spectrum graph into a plurality of frequency bands according to the frequency range relationship table;

[0093] S323. For each frequency band, calculate the sum of the energies of all frequency components in the frequency band to obtain the energy value corresponding to each frequency band;

[0094] S324. Arrange the energy values ​​of all frequency bands in order of the frequency bands to form an energy feature vector.

[0095] Specifically, the frequency range relationship table refers to a pre-constructed relationship table between the fundamental frequency vibration range and the robot arm axis, which can be stored in a table, a database or other data structure, and its purpose is to provide a basis for subsequent frequency band division.

[0096] More specifically, the fundamental frequency vibration range refers to a specific frequency interval associated with the robot arm axis, which can be determined based on the design parameters, operating conditions, or historical data analysis of the robot arm.

[0097] More specifically, the robot arm axis refers to each motion joint or degree of freedom of the robot arm, for example, it can be the rotation axis, swing axis or movement axis of the robot arm.

[0098] More specifically, the solution of this application obtains a frequency range relationship table representing the vibration characteristics of the manipulator. This relationship table reflects the vibration characteristics of different axes of the manipulator within a specific frequency range, providing a scientific basis for frequency band division of the electrical spectrum. Based on this relationship table, the frequency range of the electrical spectrum is divided into multiple frequency bands corresponding to the vibration characteristics of each axis of the manipulator, avoiding the omission of critical frequency information. For each divided frequency band, the energy sum of all frequency components within it is calculated to extract the energy characteristics of that frequency band. These frequency band energy values ​​are arranged in sequence to form an energy eigenvector. This refined frequency band division and energy extraction method based on the inherent vibration characteristics of the manipulator enables the constructed energy eigenvector to more accurately and comprehensively capture the vibration status information of the manipulator. As input for subsequent nonlinear analysis, this energy eigenvector can more effectively reflect the nonlinear characteristic changes of mechanical transmission components, thereby improving the accuracy and reliability of health status assessment of electrical automation equipment. Compared to frequency band division methods based on experience or simple rules, this solution can more accurately locate frequency components related to the manipulator's health status, improving the effectiveness of feature extraction.

[0099] Through the above technical solution, the method of this application can more accurately locate the key frequency ranges related to the vibration of each axis of the robot arm, avoiding the omission of important information. By calculating the sum of the energy within these specific frequency bands and constructing energy eigenvectors, it is possible to extract characteristic information that better reflects the actual vibration state of the robot arm. This refined feature extraction method provides high-quality data input for subsequent nonlinear analysis, thereby improving the accuracy and reliability of health status assessment of electrical automation equipment.

[0100] In some preferred implementations, the process of constructing the frequency range relationship table includes:

[0101] A1. Collect historical operating data of the robot arm under different working conditions. The historical operating data includes the type of the robot arm axis and the corresponding electrical signal data.

[0102] A2. For each type of robot arm axis, analyze the electrical signal data collected in step A1, determine the fundamental frequency vibration range of different robot arm axes under different working conditions, establish a corresponding relationship between the fundamental frequency vibration range and the robot arm axis, and form a frequency range relationship table.

[0103] Specifically, different working conditions refer to various operating states that the robotic arm may encounter during operation, such as different movement speeds, load sizes, movement trajectory types or ambient temperatures, etc., which can be achieved by using a preset test program to drive the robotic arm to operate and collect data, or by continuously collecting data during the normal production operation of the robotic arm. The purpose is to obtain vibration data covering the actual operating range of the robotic arm.

[0104] More specifically, the fundamental frequency vibration range refers to the frequency interval related to the main movement or natural frequency of the robot arm axis. It can be determined by performing spectral analysis on the collected electrical signal data to identify the main frequency components where energy is concentrated in the signal and their variation range. Its purpose is to lock the vibration frequency that is closely related to the characteristics of the robot arm axis itself.

[0105] More specifically, the solution proposed in this application analyzes the electrical signal data for each robot axis type based on historical data to determine the fundamental frequency vibration range of that axis under different operating conditions. This analysis process accurately captures the core vibration frequency characteristics of different robot axis types under different operating conditions, and then establishes a corresponding relationship between the fundamental frequency vibration range and the robot axis, forming a frequency range relationship table.

[0106] This technical solution constructs a frequency range relationship table based on the historical operating data of the robotic arm under different operating conditions, ensuring that the table accurately reflects the vibration characteristics of the robotic arm during actual operation. This makes the subsequent frequency band division based on this table more precise, and the extracted energy eigenvectors can more effectively characterize the nonlinear characteristics of the mechanical transmission components, improving the accuracy of fault diagnosis.

[0107] In some preferred implementations, the process of constructing the frequency range relationship table further includes:

[0108] A3. Using the historical operating data from step A1, calculate the energy value in each frequency band in the initial frequency range relationship table, analyze the correlation between the energy value and the health status of the robotic arm, and adjust the fundamental frequency vibration range based on the correlation analysis results to optimize the frequency range relationship table.

[0109] Specifically, based on the construction of the frequency range relationship table, this solution further improves the table's accuracy and health status reflection ability through an optimization process. Specifically, step A3 uses the historical operating data collected in step A1, which contains electrical signals under different operating conditions and possibly associated health status information, to calculate the energy value within each frequency band in the initial frequency range relationship table. By analyzing the correlation between these calculated energy values ​​and the health status of the robot arm, it is possible to identify which frequency bands have energy changes that are more closely related to health status changes. Based on the results of this correlation analysis, the initially determined fundamental frequency vibration range is adjusted to optimize the frequency range relationship table. This optimization process enables the ultimately constructed frequency range relationship table to more accurately reflect the portion of the robot arm's vibration characteristics that is highly correlated with its health status, ensuring that the subsequent extraction of nonlinear characteristic parameters based on frequency band energy can more effectively characterize the nonlinear characteristics of the mechanical transmission components and the health status of the equipment. As a result, the optimized frequency range relationship table can more accurately reflect the vibration characteristics and health status of the robot arm, improving the accuracy of subsequent health status detection.

[0110] In some preferred embodiments, in step S33, the training acquisition process of the sparse autoencoder includes:

[0111] B1. Construct a sparse autoencoder consisting of an input layer, a hidden layer, and an output layer. The number of neurons in the input layer is the same as the dimension of the energy feature vector, the number of neurons in the output layer is the same as the input layer, and the number of neurons in the hidden layer is smaller than the input layer.

[0112] B2. Initialize the weights and bias parameters of the sparse autoencoder and set the coefficient of the L1 regularization term. The L1 regularization term constrains the sparse activation characteristics of the hidden layer neurons.

[0113] B3. Input the training energy feature vector into the sparse autoencoder, calculate the reconstructed energy feature vector of the output layer, calculate the reconstruction error between the reconstructed energy feature vector and the input energy feature vector, and calculate the value of the L1 regularization term;

[0114] B4. Take the weighted sum of the reconstruction error and the L1 regularization term to obtain the total loss function value;

[0115] B5. Use the gradient descent method to optimize the weight and bias parameters of the sparse autoencoder to minimize the total loss function value;

[0116] B6. Repeat steps B3-B5 until the total loss function value converges or the maximum number of iterations is reached to obtain a trained sparse autoencoder.

[0117] Specifically, a sparse autoencoder is an unsupervised neural network used to learn a low-dimensional sparse representation of data. It can be implemented using a multilayer perceptron structure. The input layer is the part of the neural network that receives raw data. The hidden layer is the layer between the input and output layers of the neural network used to extract features. The output layer is the part of the neural network that produces results and is used here to reconstruct the input. The L1 regularization term is a technique used to constrain model complexity. By penalizing the sum of the absolute values ​​of the model parameters, it encourages the model parameters to approach zero, thereby achieving sparse activation of hidden layer neurons.

[0118] More specifically, the solution of the present application constructs a sparse autoencoder with a specific structure, in which the number of neurons in the input layer matches the dimension of the energy feature vector to be processed, the number of neurons in the output layer is the same as that in the input layer to achieve reconstruction, and the number of neurons in the hidden layer is less than that in the input layer, thus naturally having the ability of dimensionality reduction.

[0119] More specifically, during the training process, the model parameters are initialized, and an L1 regularization term is introduced to constrain the activation of hidden layer neurons. The strength of this constraint is controlled by a set coefficient, encouraging the model to learn sparse feature representations. The training data, namely the energy feature vector, is input into the model. The model attempts to reconstruct the input and calculates the reconstruction error between the reconstructed output and the original input. At the same time, the value of the L1 regularization term of the hidden layer activation is calculated. The reconstruction error and the value of the L1 regularization term are weighted and summed to form the total loss function value, which comprehensively measures the reconstruction ability and sparsity of the model. Using the gradient descent method, the model iteratively adjusts the weights and bias parameters according to the gradient information of the total loss function value to minimize the total loss function value. This iterative process continues until the total loss function value changes slightly or reaches the preset maximum number of iterations, at which point the model training is complete.

[0120] More specifically, a trained sparse autoencoder can map high-dimensional energy feature vectors to a low-dimensional hidden layer space. This low-dimensional representation is the nonlinear feature parameter after dimensionality reduction. These parameters not only reduce dimensionality, reducing the computational burden of subsequent processing, but also, due to the nonlinear activation and L1 regularization constraints of the sparse autoencoder, the extracted features have nonlinear characteristics and sparsity, better capturing the essential information related to the health status of the equipment in the data. This training process, combined with the aforementioned step of extracting energy feature vectors, enables the energy features extracted from the original electrical signal to be effectively compressed and refined, providing more efficient and discriminative input data for subsequent health status detection based on nonlinear feature parameters.

[0121] The method of this application uses a sparse autoencoder to train and reduce the dimensionality of high-dimensional energy feature vectors, effectively reducing the dimensionality of the feature vectors and reducing data redundancy. At the same time, the sparse autoencoder can extract low-dimensional feature parameters with nonlinear characteristics. These parameters can better reflect the nonlinear characteristic changes of the mechanical transmission components of electrical automation equipment. This significantly reduces the computational complexity of the subsequent health status detection algorithm and improves processing speed and efficiency. In addition, the extracted low-dimensional nonlinear feature parameters have stronger discriminative power, improving the accuracy of health status detection.

[0122] In some preferred embodiments, step S2 includes:

[0123] S21. Establishing a mapping relationship between the robot arm control information and the timestamp of the electrical signal;

[0124] S22, analyzing the robot arm control information to obtain and identify a control instruction subsequence of a repeated motion segment, and determining its start and end timestamps;

[0125] S23. Extract the electrical signal segments corresponding to the start and end timestamps of the repeated motion segment control instruction subsequence according to the mapping relationship as the electrical signal segments of the repeated motion segment.

[0126] Specifically, analyzing the control information of the robotic arm to obtain and identify the control instruction subsequence of repetitive motion segments refers to parsing and processing the control instruction data stream received or executed by the robotic arm to identify a series of continuous instructions that constitute a specific repetitive motion pattern. This can be achieved by parsing the teaching program, identifying specific motion instruction patterns, or analyzing joint trajectory planning data.

[0127] More specifically, the solution of the present application solves the time synchronization problem between different data sources by establishing a mapping relationship between the timestamps of the robot arm control information and the electrical signal, ensuring the accurate alignment of the control information and the electrical signal in time. On this basis, by analyzing the robot arm control information, the control instruction subsequences that constitute the repetitive motion segments are accurately identified, and the starting and ending points of these instruction subsequences in time are determined. Subsequently, using the previously established timestamp mapping relationship, the time boundaries of the control instruction subsequences are converted into the time boundaries of the electrical signal, so that electrical signal segments that are highly corresponding to the repetitive motion segments can be accurately extracted from the electrical signal. This method avoids signal segment deviations caused by time asynchrony or inaccurate motion segment identification, ensuring the accuracy of the extracted data. Applying this accurately extracted electrical signal segment to subsequent nonlinear analysis can more effectively capture the subtle nonlinear characteristic changes generated by mechanical transmission components during repetitive motion, thereby improving the reliability and sensitivity of the entire electrical automation equipment health status detection method.

[0128] In some preferred embodiments, step S4 includes:

[0129] S41, obtaining predetermined parameter benchmark data, the parameter benchmark data including benchmark values ​​of multiple nonlinear characteristic parameters and corresponding weight coefficients;

[0130] S42, comparing the nonlinear characteristic parameters with the reference values ​​in the parameter reference data one by one, and calculating the difference between each nonlinear characteristic parameter and the corresponding reference value;

[0131] S43, performing weighted processing on the difference calculated in step S42 according to the weight coefficient to obtain a weighted change of each nonlinear characteristic parameter;

[0132] S44. Accumulate the weighted changes of all nonlinear characteristic parameters to obtain the changes of the nonlinear characteristic parameters.

[0133] Specifically, parameter benchmark data refers to a reference data set used to measure the health status of electrical automation equipment. It can use pre-collected operating data of the equipment in a normal health state, and extract nonlinear characteristic parameters from it, calculate their average value or statistical distribution as a benchmark value, and at the same time, according to the sensitivity or importance of different nonlinear characteristic parameters to the health status of the equipment, determine the corresponding weight coefficient through expert experience, historical fault data analysis or machine learning methods. Its purpose is to provide a standard reference for evaluating the current status of the equipment and a measure of the importance of different characteristic parameters; the weight coefficient refers to a numerical value assigned to each nonlinear characteristic parameter to indicate the relative importance of the parameter in evaluating the health status of the equipment. It can be expressed by a numerical value between 0 and 1 or other proportional numerical values. Its purpose is to distinguish the contribution of different nonlinear characteristic parameters to the health status of the equipment, and can be set according to usage requirements.

[0134] More specifically, the weighted change refers to the change in the value of each nonlinear characteristic parameter after weighted processing, which can be expressed as the product of the difference and the weight coefficient. Its purpose is to quantify the weighted impact of each nonlinear characteristic parameter change on the health status of the equipment.

[0135] More specifically, the solution of the present application obtains parameter baseline data containing a baseline value and a weight coefficient, compares the currently extracted nonlinear characteristic parameter with the baseline value to obtain a difference, then weights these differences according to the weight coefficient to obtain a weighted change in each parameter, and finally accumulates all weighted changes to obtain a comprehensive nonlinear characteristic parameter change. It is precisely because of the introduction of the weight coefficient and the weighted processing of the changes in different nonlinear characteristic parameters that the nonlinear characteristic parameter changes finally obtained can more accurately reflect the overall health status of the equipment, because important parameter changes will be amplified and the impact of minor parameter changes will be reduced. This more accurate nonlinear characteristic parameter change is used in the subsequent health status assessment step, thereby improving the assessment accuracy and reliability of the entire detection method.

[0136] Second, please refer to Figure 3 and Figure 4 Some embodiments of the present application further provide a detection system for electrical automation equipment, which is applied to electrical automation equipment having a robotic arm. The system includes:

[0137] The acquisition module 201 is used to acquire the electrical signals of the robot arm and obtain the corresponding robot arm control information;

[0138] a slicing module 202 for identifying electrical signal segments corresponding to repetitive motion segments of the robotic arm from the electrical signal based on the robotic arm control information;

[0139] A feature extraction module 203 is used to perform nonlinear analysis on the electrical signal segment and extract nonlinear feature parameters, which represent the nonlinear characteristics of the mechanical transmission component;

[0140] Comparison module 204, used to compare the nonlinear characteristic parameter with predetermined parameter reference data to obtain a variation of the nonlinear characteristic parameter;

[0141] The detection module 205 is used to detect the current health status of the electrical automation equipment according to the variation of the nonlinear characteristic parameter and a preset health threshold.

[0142] The electrical automation equipment detection system described in this application, based on robotic arm control information, effectively separates signal segments related to the repetitive motion of mechanical transmission components from dynamic electrical signals subject to interference from various factors. It also utilizes a nonlinear analysis system to sensitively capture nonlinear characteristic changes caused by early wear. Compared to traditional linear analysis systems, this system can more accurately and earlier identify early wear characteristics of mechanical transmission components online, thereby enabling early warning and assessment of the health status of electrical automation equipment, preventing equipment maintenance from lagging behind actual wear progression.

[0143] In some preferred embodiments, the system further comprises:

[0144] The prompt module 206 is used to generate an abnormal status prompt when the current health status is abnormal.

[0145] In addition, the units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0146] Furthermore, the functional modules in each embodiment of the present application can be integrated together to form an independent part, or each module can exist independently, or two or more modules can be integrated to form an independent part.

[0147] In this document, relational terms such as first and second, etc. are used merely to distinguish one entity or operation from another entity or operation, but do not necessarily require or imply any actual relationship or order between these entities or operations.

[0148] The above description is merely an embodiment of the present application and is not intended to limit the scope of protection of the present application. For those skilled in the art, various modifications and variations of the present application are possible. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application shall be included in the scope of protection of the present application.

Claims

1. A detection method for electrical automation equipment, applied to electrical automation equipment with a robotic arm, characterized in that: The method comprises the following steps: S1. Collect the electrical signal of the robot arm and obtain the corresponding robot arm control information; S2. identifying, from the electrical signal, an electrical signal segment corresponding to a repetitive motion segment of the robotic arm based on the robotic arm control information; S3, performing nonlinear analysis on the electrical signal segment to extract nonlinear characteristic parameters, wherein the nonlinear characteristic parameters represent the nonlinear characteristics of the mechanical transmission component; S4, comparing the nonlinear characteristic parameter with predetermined parameter reference data to obtain a variation of the nonlinear characteristic parameter; S5. Detecting the current health status of the electrical automation equipment based on the nonlinear characteristic parameter change and a preset health threshold; Step S3 includes: S31, performing short-time Fourier transform on the electrical signal segment to obtain an electrical spectrum diagram; S32. Divide the frequency range into multiple frequency bands based on the vibration characteristics of the robotic arm, and calculate the energy value in each frequency band according to the electrical spectrum diagram to form an energy feature vector; S33, using a pre-trained sparse autoencoder to perform dimensionality reduction processing on the energy feature vector to perform nonlinear analysis, and obtain nonlinear feature parameters after dimensionality reduction; Step S4 includes: S41, obtaining predetermined parameter benchmark data, wherein the parameter benchmark data includes benchmark values ​​of multiple nonlinear characteristic parameters and corresponding weight coefficients; S42, comparing the nonlinear characteristic parameters with the reference values ​​in the parameter reference data one by one, and calculating the difference between each nonlinear characteristic parameter and the corresponding reference value; S43, performing weighted processing on the difference calculated in step S42 according to the weight coefficient to obtain a weighted variation of each nonlinear characteristic parameter; S44. Accumulate the weighted changes of all nonlinear characteristic parameters to obtain the changes of the nonlinear characteristic parameters.

2. The detection method of electrical automation equipment according to claim 1, characterized in that: The electrical signal includes a current signal and / or a voltage signal of a servo motor of the robotic arm.

3. The detection method of electrical automation equipment according to claim 1, characterized in that: The manipulator control information includes one or more of a manipulator control instruction, a joint encoder signal, and a motor torque instruction.

4. The detection method of electrical automation equipment according to claim 1, characterized in that: Step S32 includes: S321. Obtain a frequency range relationship table representing the vibration characteristics of the robotic arm, where the frequency range relationship table is a pre-constructed relationship table between the fundamental frequency vibration range and the robotic arm axis; S322, dividing the frequency range in the electrical spectrum graph into a plurality of frequency bands according to the frequency range relationship table; S323. For each frequency band, calculate the sum of the energies of all frequency components in the frequency band to obtain the energy value corresponding to each frequency band; S324. Arrange the energy values ​​of all frequency bands in order of frequency bands to form an energy feature vector.

5. The detection method of electrical automation equipment according to claim 4, characterized in that: The process of constructing the frequency range relationship table includes: A1. Collect historical operating data of the robotic arm under different working conditions, including the types of the robotic arm axes and corresponding electrical signal data; A2. For each type of robotic arm axis, analyze the electrical signal data collected in step A1 to determine the fundamental frequency vibration range of different robotic arm axes under different working conditions, establish a corresponding relationship between the fundamental frequency vibration range and the robotic arm axis, and form a frequency range relationship table.

6. The detection method of electrical automation equipment according to claim 1, characterized in that: In step S33, the training acquisition process of the sparse autoencoder includes: B1. Construct a sparse autoencoder consisting of an input layer, a hidden layer, and an output layer. The number of neurons in the input layer is the same as the dimension of the energy feature vector, the number of neurons in the output layer is the same as the input layer, and the number of neurons in the hidden layer is smaller than the input layer. B2. Initialize the weights and bias parameters of the sparse autoencoder and set the coefficient of the L1 regularization term; B3. Input the training energy feature vector into the sparse autoencoder, calculate the reconstructed energy feature vector of the output layer, calculate the reconstruction error between the reconstructed energy feature vector and the input energy feature vector, and calculate the value of the L1 regularization term; B4. Taking a weighted sum of the reconstruction error and the value of the L1 regularization term to obtain a total loss function value; B5. Use the gradient descent method to optimize the weight and bias parameters of the sparse autoencoder to minimize the total loss function value; B6. Repeat steps B3-B5 until the total loss function value converges or the maximum number of iterations is reached to obtain a trained sparse autoencoder.

7. The detection method of electrical automation equipment according to claim 1, characterized in that: Step S2 includes: S21. Establishing a mapping relationship between the robot arm control information and the timestamp of the electrical signal; S22, analyzing the robot arm control information to obtain and identify a control instruction subsequence of a repeated motion segment, and determining its start and end timestamps; S23. Extract, according to the mapping relationship, electrical signal segments corresponding to the start and end timestamps of the repeated motion segment control instruction subsequence as electrical signal segments of the repeated motion segment.

8. A detection system for electrical automation equipment, applied to electrical automation equipment having a robotic arm, for executing the detection method for electrical automation equipment according to any one of claims 1 to 7, characterized in that: The system comprises: The acquisition module is used to collect the electrical signals of the robotic arm and obtain the corresponding robotic arm control information; a slicing module, configured to identify, from the electrical signal, electrical signal segments corresponding to repetitive motion segments of the robotic arm based on the robotic arm control information; A feature extraction module is used to perform nonlinear analysis on electrical signal segments and extract nonlinear feature parameters, which represent the nonlinear characteristics of mechanical transmission components; A comparison module, configured to compare the nonlinear characteristic parameter with predetermined parameter reference data to obtain a variation of the nonlinear characteristic parameter; The detection module is used to detect the current health status of the electrical automation equipment according to the change amount of the nonlinear characteristic parameter and a preset health threshold.

Images