A method and system for acquiring and analyzing power supply operation data of industrial robots

Abstract

This application provides a method and system for acquiring and analyzing power supply operation data for industrial robots. Applied to the field of industrial robot technology, the method synchronously acquires electrical parameter signals and performs time-frequency domain transformation on these signals to obtain multi-channel time-frequency feature vectors. Next, independent component decomposition is performed on the multi-channel time-frequency feature vectors to separate the first signal component characterizing the internal operating state of the power supply. Simultaneously, real-time motion control commands from the industrial robot are acquired, and a mapping relationship is established between the motion control commands and the fundamental energy in the first signal component to identify load fluctuation components. Then, the load fluctuation component is removed from the first signal component, and the background degradation signal characterizing the physical properties of the components is extracted. Finally, based on the evolution trend of the background degradation signal over time, the health assessment result of the power supply system is determined. This method can effectively distinguish between changes in the power supply's own operating state and interference from the external environment, avoiding misjudgments and missed detections.

Description

Technical Field

[0001] This application relates to the field of industrial robot technology, and in particular to a method and system for acquiring and analyzing power supply operation data of industrial robots. Background Technology

[0002] In industrial automated production, monitoring the power status of robots is crucial for ensuring production line stability. Traditional methods establish static data benchmarks based on fixed tasks, effectively identifying anomalies, such as high motor current indicating mechanical problems. However, as production becomes more flexible and diversified, robots frequently switch tasks (e.g., handling workpieces of varying weights), causing drastic fluctuations in power data. Continuing to use static standards will misjudge normal high loads as faults, generating numerous false alarms; while relaxing thresholds may miss subtle fault signs. To address this issue, existing technologies correlate power data with the robot's real-time task information and compare it with corresponding historical data, successfully distinguishing normal fluctuations caused by task differences.

[0003] However, when this method is used in a workshop, especially when large stamping, spot welding and other equipment are started and stopped, it will inject electromagnetic noise such as voltage spikes and drops into the power grid. These interferences will be recorded by high-precision sensors, and their waveforms are very similar to the early failure characteristics of the power supply, such as the ripple generated by the aging of the filter capacitor.

[0004] Because existing technologies cannot distinguish whether an anomaly originates from its own fault or external power grid disturbance after linking power data with the robot's real-time task information, data analysis systems are prone to misjudging power grid noise as power failure and issuing false alarms, or submerging real component aging signals in strong interference and failing to report them, ultimately causing sudden shutdowns and resulting in longer production line interruptions.

[0005] To address the aforementioned issues, existing technologies urgently need improvement. Summary of the Invention

[0006] In view of the shortcomings of the prior art, this application provides a method and system for acquiring and analyzing power supply operation data of industrial robots, which has the advantages of being able to accurately distinguish between changes in the power supply's own operating status and external environmental interference, and avoiding misjudgment and missed reporting.

[0007] A first aspect is a method for collecting and analyzing power supply operation data of an industrial robot, the method comprising the following steps: S1: Synchronously acquire electrical parameter signals of multiple power input nodes and power output nodes within the power system; S2: Perform time-frequency domain transformation processing on the multiple electrical parameter signals to extract the frequency distribution characteristics of the dynamic evolution of each electrical parameter signal on the time axis, and obtain a multi-channel time-frequency feature vector; S3: Perform independent component decomposition on the multi-channel time-frequency feature vector to separate the first signal component that characterizes the internal operating state of the power supply. S4: Obtain the real-time motion control command of the industrial robot and establish the mapping relationship between the motion control command and the fundamental energy in the first signal component to identify the load fluctuation component caused by the robot's action. S5: Remove the load fluctuation component from the first signal component and extract the background degradation signal that characterizes the physical properties of the component; S6: Based on the evolution trend of the background degradation signal in the time series, determine the health assessment result of the power system.

[0008] Furthermore, step S4 includes: S41: Real-time acquisition of the torque output value or speed setpoint of the joint motor as the motion control command; S42: Perform multivariate regression analysis on the motion control command and the fundamental energy of the first signal component within the same time window to construct a mapping function; S43: Calculate the fundamental energy change expected by the current motion control command according to the mapping function, and identify the signal component corresponding to the change as the load fluctuation component.

[0009] Furthermore, step S5 includes: S51: Subtract the time-domain waveform of the load fluctuation component from the first signal component in the time domain to obtain the residual signal; S52: Perform adaptive filtering on the residual signal to eliminate the fundamental frequency fluctuation component remaining due to mapping error, and output a smoothed background degradation signal.

[0010] Furthermore, step S6 includes: S61: The background degradation signal is segmented according to a fixed time window, and the amplitude sequence of multiple characteristic frequencies related to the aging of power supply components within each window is extracted; S62: Fit the exponential growth curves of each amplitude sequence and calculate its growth rate; S63: When the growth rate of at least one characteristic frequency exceeds a preset rate threshold, a comprehensive health index is calculated by combining the weighting coefficients of each characteristic frequency, and a health level is output according to the interval in which the comprehensive health index is located.

[0011] Furthermore, step S1 includes: S11: The electrical parameter signals of the power input node and the power output node are synchronously captured using a sampling frequency of not less than 100kHz to obtain an original sampling sequence containing the transient voltage spike characteristics of the external power grid and the weak high-frequency ripple characteristics generated by the aging of internal components of the power supply. The original sampling sequence is the digital expression of the electrical parameter signals.

[0012] Furthermore, step S1 also includes: S12: The central controller sends a synchronous sampling trigger signal to the analog-to-digital converter of the sensor deployed at the power input node and the power output node to ensure the physical synchronization of the sampling actions of each channel; S13: Associate each acquired data point with a nanosecond-level timestamp, and perform phase compensation on the electrical parameter signals of different channels based on the timestamps to eliminate the differences in signal propagation delay at different points.

[0013] Furthermore, in step S13, associating a nanosecond-level timestamp with each collected data point includes the following steps: S131: Send a global synchronization pulse to all sensors; S132: Record the local count value of the global synchronization pulse received by each of the sensors, and calculate the clock drift coefficient of each sensor by combining it with the standard clock frequency of the central controller; S133: Perform linear compensation on the original sampling time of each of the sensors according to the clock drift coefficient to generate the nanosecond-level timestamp under a unified clock reference.

[0014] Furthermore, step S2 includes: S21: Use short-time Fourier transform to perform frame-by-frame windowing processing on each of the electrical parameter signals to obtain the spectral distribution within each time window; S22: Arrange the spectral distribution of all time windows within the same channel in chronological order to form the time-frequency characteristic sequence of that channel; S23: Combine the time-frequency feature sequences of all channels into the multi-channel time-frequency feature vector.

[0015] Furthermore, step S3 includes: S31: Extract multiple statistically independent source signal vectors from the multi-channel time-frequency feature vector using the independent component analysis algorithm; S32: Calculate the energy attenuation ratio of each source signal vector between the power input node and the power output node, and determine the source signal vector with an energy attenuation ratio greater than a preset threshold and whose time-frequency characteristics are instantaneous wideband pulses as the first signal component.

[0016] Secondly, an industrial robot power supply operation data acquisition and analysis system, the system being used to implement the steps of any of the methods described above, the system comprising: Acquisition module: Synchronously acquires electrical parameter signals from multiple power input nodes and power output nodes within the power system; Time-frequency transformation module: Performs time-frequency domain transformation processing on the multiple electrical parameter signals to extract the frequency distribution characteristics of the dynamic evolution of each electrical parameter signal on the time axis, and obtains a multi-channel time-frequency feature vector; Separation module: Performs independent component decomposition on the multi-channel time-frequency feature vector to separate the first signal component characterizing the internal operating state of the power supply; Mapping relationship establishment module: acquires the real-time motion control commands of the industrial robot and establishes a mapping relationship between the motion control commands and the fundamental energy in the first signal component to identify the load fluctuation component caused by the robot's actions; Extraction module: Removes the load fluctuation component from the first signal component and extracts the background degradation signal characterizing the physical properties of the component; Evaluation module: Based on the evolution trend of the background degradation signal over time, determine the health assessment result of the power system.

[0017] Beneficial Effects: This application proposes a method and system for acquiring and analyzing power supply operation data of an industrial robot. It synchronously acquires electrical parameter signals from multiple power input and output nodes within the power supply system, performs time-frequency domain transformation on these signals, extracts dynamically evolving frequency distribution characteristics, and obtains multi-channel time-frequency feature vectors. Next, it performs independent component decomposition on the multi-channel time-frequency feature vectors to separate the first signal component characterizing the internal operating state of the power supply. Simultaneously, it acquires real-time motion control commands from the industrial robot and establishes a mapping relationship between the motion control commands and the fundamental energy in the first signal component to identify load fluctuation components caused by robot movements. Then, it removes the load fluctuation component from the first signal component and extracts the background degradation signal characterizing the physical properties of the components. Finally, based on the evolution trend of the background degradation signal over time, it determines the health assessment result of the power supply system. This method can effectively distinguish between changes in the power supply's own operating state and interference from the external environment, avoiding misjudgments and missed detections. Attached Figure Description

[0018] Figure 1 This is a flowchart of a method for collecting and analyzing power supply operation data of an industrial robot proposed in this application.

[0019] Figure 2 This is a structural diagram of an industrial robot power supply operation data acquisition and analysis system proposed in this application.

[0020] Figure 3This is a schematic diagram of an industrial robot power supply operation data acquisition and analysis system proposed in this application.

[0021] Labeling Explanation: 201. Acquisition Module; 202. Time-Frequency Transformation Module; 203. Separation Module; 204. Mapping Relationship Establishment Module; 205. Extraction Module; 206. Evaluation Module. Detailed Implementation

[0022] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. The components of the embodiments of this application described and marked in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely represents selected embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0023] It should be noted that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. Furthermore, in the description of this application, terms such as "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0024] Traditional methods for monitoring the power status of industrial robots typically establish static data benchmarks based on fixed task modes. This approach is effective in identifying anomalies when the robot is performing single, repetitive tasks. However, as modern production models become more flexible and diversified, industrial robots need to frequently switch tasks, leading to drastic fluctuations in power operation data and generating numerous false alarms or missed alarms. Furthermore, the start-up and shutdown of high-power equipment such as large stamping and spot welding machines inject strong electromagnetic noise, including voltage spikes and drops, into the power grid. The waveform characteristics of this noise are, in some cases, extremely similar to early power supply failure characteristics, causing data analysis systems to easily misinterpret grid noise as power supply faults, resulting in false alarms. Alternatively, genuine aging signals from power components may be masked by strong external electromagnetic interference, leading to missed alarms.

[0025] For this, please refer to Figure 1 This application proposes a method for acquiring and analyzing power supply operation data of industrial robots, the method including the following steps: S1: Synchronously acquire electrical parameter signals of multiple power input nodes and power output nodes within the power system; S2: Perform time-frequency domain transformation on the multi-channel electrical parameter signals to extract the frequency distribution characteristics of the dynamic evolution of each electrical parameter signal on the time axis, and obtain the multi-channel time-frequency feature vector; S3: Perform independent component decomposition on the multi-channel time-frequency feature vector to separate the first signal component that characterizes the internal operating state of the power supply. S4: Obtain the real-time motion control commands of the industrial robot and establish the mapping relationship between the motion control commands and the fundamental energy in the first signal component in order to identify the load fluctuation component caused by the robot's actions. S5: Remove the load fluctuation component from the first signal component and extract the background degradation signal that characterizes the physical properties of the component; S6: Based on the evolution trend of the background degradation signal in the time series, determine the health assessment results of the power system.

[0026] In step S1, electrical parameter signals refer to physical quantities that reflect the electrical characteristics of a power supply system, such as voltage, current, and power. Within a power supply system, there are multiple power input nodes and multiple power output nodes. These input and output nodes are located in different physical locations, and their respective electrical parameter signals also differ. By synchronously acquiring the electrical parameter signals of multiple power input and output nodes within the power supply system, a complete dataset reflecting the overall operating status of the system can be obtained under the same time reference.

[0027] To synchronously acquire multiple electrical parameter signals, multiple independent sensors can be deployed. Each sensor is responsible for acquiring the electrical parameter signal from a power input node or power output node, and converts the analog signal into a digital signal through its own analog-to-digital converter. These sensors can be pre-configured to operate at the same sampling frequency and sampling period to achieve synchronous acquisition. Alternatively, a multi-channel data acquisition card can be used. This card has multiple input channels, each connected to a power input node or power output node. The analog-to-digital converter inside the acquisition card can sample the signals from all channels in parallel, thus achieving physical synchronization. For example, in an industrial robot power system, current and voltage sensors can be deployed at the AC input, DC output, and the input / output terminals of key power conversion modules (such as rectifiers and inverters) to acquire the current and voltage signals at these key points as electrical parameter signals.

[0028] In some embodiments of this application, step S1 only proposes to simultaneously acquire electrical parameter signals from multiple power input nodes and power output nodes within the power system. However, if the sampling frequency is insufficient or the sampling accuracy is low, it may be unable to effectively capture the weak high-frequency ripple characteristics generated by the aging of internal components and the transient voltage spike characteristics of the external power grid, thereby affecting the accuracy of subsequent health assessments. Therefore, this application further proposes to optimize the acquisition method of the aforementioned electrical parameter signals to ensure that the acquired data can comprehensively reflect the operating status of the power supply.

[0029] Furthermore, step S1 includes: S11: The electrical parameter signals of the power input node and the power output node are synchronously captured using a sampling frequency of not less than 100kHz to obtain the original sampling sequence containing the transient voltage spike characteristics of the external power grid and the weak high-frequency ripple characteristics generated by the aging of internal components of the power supply. The original sampling sequence is a digital expression of the electrical parameter signal.

[0030] Specifically, synchronously capturing electrical parameter signals using a sampling frequency of at least 100kHz refers to digitally acquiring electrical parameters such as voltage and current at the power input and output nodes using a high-precision analog-to-digital converter at a rate of at least 100kHz. The purpose is to ensure the capture of all important dynamic changes during power system operation, including rapid transient events and high-frequency noise. External grid transient voltage spikes refer to short-duration, high-amplitude voltage or current anomalies caused by external factors such as grid fluctuations, switching operations, or lightning strikes; these spikes may impact power supply components. Weak high-frequency ripple characteristics generated by the aging of internal power supply components refer to small, high-frequency periodic fluctuations superimposed on the normal operating signal, caused by internal aging mechanisms such as increased capacitor ESR (equivalent series resistance), inductor saturation, or degraded semiconductor device performance. The original sampling sequence can be understood as a set of digital samples continuously arranged on the time axis after analog-to-digital conversion; it is a precise expression of the electrical parameter signals in the digital domain.

[0031] By using these raw sampling sequences containing rich details as digital representations of electrical parameter signals, a high-quality raw data foundation is provided for subsequent time-frequency domain analysis and background degradation signal extraction, thereby significantly improving the sensitivity and accuracy of power supply health assessment.

[0032] Furthermore, in real-world industrial environments, due to the dispersed locations of sensors, differences in signal transmission paths, and minute deviations in the response time of analog-to-digital converters, simple synchronous acquisition may not guarantee precise alignment of data from each channel at the nanosecond or even picosecond level. This minute timing deviation or phase lag, especially when acquiring high-frequency signals or requiring fine time-frequency analysis, can introduce errors, affecting the accuracy of subsequent signal processing and thus reducing the reliability of identifying the internal operating status of the power supply.

[0033] Furthermore, step S1 also includes: S12: The central controller sends a synchronous sampling trigger signal to the analog-to-digital converter of the sensor deployed at the power input node and the power output node to ensure the physical synchronization of the sampling actions of each channel; S13: Associate a nanosecond-level timestamp with each acquired data point, and perform phase compensation on the electrical parameter signals of different channels based on the timestamp to eliminate the difference in signal propagation delay at different points.

[0034] Specifically, in step S12, the central controller is responsible for generating and distributing a unified synchronization sampling trigger signal. This signal is sent to all sensors participating in data acquisition, especially their internal analog-to-digital converters. In this way, it can be ensured that all analog-to-digital converters start their sampling operation almost simultaneously upon receiving the trigger signal. This physical-level synchronization mechanism aims to minimize the initial sampling timing deviation caused by differences in the startup times of different devices.

[0035] In step S13, each acquired data point is associated with a nanosecond-level timestamp, meaning that each digitized electrical parameter signal sample carries an extremely precise time stamp. These timestamps can be generated by a high-precision clock inside the sensor or synchronized by a central controller. In practical applications, based on these nanosecond-level timestamps, fine phase compensation can be performed on the electrical parameter signals of different channels. For example, if it is known that the signal of a certain channel has a slight propagation delay due to a long transmission distance, this delay can be calculated using the timestamp information, and the signal of that channel can be shifted forward or backward on the time axis during the data processing stage, thereby eliminating the difference in propagation delay at different points and achieving precise alignment of signals from each channel.

[0036] The solution proposed in this application effectively solves the timing deviation problem that may exist in the acquisition of multi-channel electrical parameter signals by introducing a synchronous sampling trigger signal and a nanosecond-level timestamp and phase compensation mechanism.

[0037] Specifically, in the above method, the step of associating each collected data point with a nanosecond-level timestamp can be further refined into the following sub-steps: S131: Send a global synchronization pulse to all sensors; S132: Record the local count value of the global synchronization pulse received by each sensor, and calculate the clock drift coefficient of each sensor by combining it with the standard clock frequency of the central controller. S133: Linearly compensate the original sampling time of each sensor according to the clock drift coefficient to generate a nanosecond-level timestamp under a unified clock reference.

[0038] Step S131 aims to provide a unified, high-precision time reference point for all distributed sensors. This global synchronization pulse can be understood as a periodic time calibration signal, ensuring that each sensor can use it as a time synchronization reference when it receives the pulse.

[0039] Furthermore, in step S132, since the internal clocks of each sensor may have slight frequency deviations or drifts, these local count values ​​will differ from the standard clock frequency of the central controller. By comparing these differences, the clock drift coefficient of each sensor can be accurately calculated. This clock drift coefficient characterizes how fast or slow the sensor's local clock is relative to the standard clock.

[0040] Specifically, the method for calculating the clock drift coefficient is as follows: first, use the standard time interval between two consecutive synchronization signals. Incremental local count of sensor Combined with the sensor's nominal frequency The actual frequency ratio is obtained as follows: Clock drift coefficient Then it is Among them, the sensor local count increment This is the difference between the local counter counts when the sensor receives electrical signal parameters twice consecutively. Standard time interval. This is the time difference recorded by a standard clock when the sensor receives electrical signal parameters twice consecutively. Sensor nominal frequency. The clock drift factor is determined by the sensor model and is usually indicated at the factory in the instruction manual or on the housing. A positive value indicates that the sensor's local clock is faster than the standard clock; a negative value indicates that it is slower.

[0041] Therefore, in step S133, the compensation formula for linearly compensating the original sampling time of each sensor is: Compensated time = Original sampling time - (Original sampling time - Reference start time) * Clock drift coefficient.

[0042] This solution actively measures and corrects the clock drift of each sensor, ensuring high accuracy and consistency of the generated nanosecond-level timestamps, providing a reliable time reference for subsequent precise phase compensation. Thus, even in complex industrial environments, fine-grained time alignment of multiple electrical parameter signals can be achieved.

[0043] In step S2, time-frequency domain transformation is a signal processing technique that can simultaneously analyze the characteristics of a signal in time and frequency, such as short-time Fourier transform. This processing can reveal the frequency components of a signal that change over time, thereby providing a more comprehensive understanding of the dynamic behavior of the signal.

[0044] Multi-channel time-frequency feature vectors are feature sets obtained by transforming multiple electrical parameter signals in the time-frequency domain. They integrate time-frequency information from different measurement points, providing rich data for subsequent signal separation and analysis.

[0045] Specifically, step S2 includes: S21: Use short-time Fourier transform to perform frame-by-frame windowing processing on each electrical parameter signal to obtain the spectral distribution within each time window; S22: Arrange the spectral distribution of all time windows within the same channel in chronological order to form the time-frequency characteristic sequence of that channel; S23: Combine the time-frequency feature sequences of all channels into a multi-channel time-frequency feature vector.

[0046] In step S21, the short-time Fourier transform (SFT) is used to perform frame-based windowing processing on each electrical parameter signal. This involves dividing the continuous electrical parameter signal into a series of overlapping or non-overlapping short time frames and applying a window function, such as a Hanning window, to each time frame to reduce spectral leakage. A SFT is then performed on each windowed frame. This transform calculates the frequency components of the signal within that frame and their corresponding complex values. Amplitude and phase information are extracted from the complex values ​​to obtain the spectral distribution within that time window. This processing method aims to capture the local frequency characteristics of the signal within a short time.

[0047] Furthermore, in step S22, for each independent electrical parameter signal acquisition channel, the spectral distributions obtained within different time windows are arranged sequentially to construct a two-dimensional data structure. This structure can intuitively show how the frequency components of the channel signal dynamically evolve over time. This time-frequency characteristic sequence can reflect information such as transient changes and frequency drift of the signal.

[0048] Therefore, in step S23, by integrating the respective time-frequency characteristic sequences formed by the electrical parameter signals from multiple power input nodes and power output nodes within the power system, a comprehensive multi-channel time-frequency characteristic vector can be constructed. This vector contains information on the dynamic frequency evolution of the power system at different physical locations and time points, providing a comprehensive data foundation for subsequent signal separation and health assessment.

[0049] In step S3, independent component decomposition is a blind source separation technique designed to separate statistically independent source signals from mixed signals, such as independent component analysis (ICA). The first signal component, after independent component decomposition, is identified as an independent signal component characterizing the internal operating state of the power supply. It eliminates the influence of external interference and robot load fluctuations, and more directly reflects the health status of the power supply itself.

[0050] Specifically, in the above-mentioned method for collecting and analyzing power supply operation data of industrial robots, step S3 includes: S31: Extract multiple statistically independent source signal vectors from the multi-channel time-frequency feature vector using the independent component analysis algorithm; S32: Calculate the energy attenuation ratio of each source signal vector between the power input node and the power output node, and determine the source signal vector with an energy attenuation ratio greater than a preset threshold and whose time-frequency characteristics are instantaneous wideband pulses as the first signal component.

[0051] Independent Component Analysis (ICA) is applied to the multi-channel time-frequency eigenvectors to separate the mixed signal into statistically independent source signals. These source signal vectors can be understood as independent representations of different physical processes or noise sources within the power supply system. ICA effectively decouples signals related to the power supply's internal operating state from other interference signals, laying the foundation for subsequent analysis.

[0052] In step S32, the energy attenuation ratio of each source signal vector between the power input node and the power output node is calculated. This energy attenuation ratio reflects the degree of energy loss during signal transmission within the power supply system. Typically, signals caused by degradation or failure of internal components within the power supply exhibit significant energy attenuation within the power supply. Simultaneously, the time-frequency characteristics of each source signal vector are evaluated, with signals exhibiting instantaneous wideband pulses used as one of the screening criteria. Instantaneous wideband pulse characteristics are often associated with internal fault phenomena such as partial discharge, electric arcing, or switching transients. Therefore, only when the energy attenuation ratio of a source signal vector is greater than a preset threshold, and its time-frequency characteristics conform to the characteristics of an instantaneous wideband pulse, will it be determined as the first signal component.

[0053] The solution proposed in this application utilizes the independent component analysis algorithm to process the multi-channel time-frequency feature vectors, which can effectively decompose the complex mixed signals inside the power supply system into multiple statistically independent source signal vectors.

[0054] In step S4, the motion control command is issued by the industrial robot controller to guide the robot to perform specific actions.

[0055] Fundamental frequency energy refers to the energy contained in the fundamental frequency component of a signal, and it is usually closely related to the output power of the power supply and the load size.

[0056] The load fluctuation component refers to the change in the fundamental power supply energy caused by the actions of an industrial robot, reflecting the changes in the power demand of the robot when performing tasks.

[0057] Specifically, step S4 includes: S41: Real-time acquisition of the torque output value or speed setpoint of the joint motor as motion control command; S42: Perform multivariate regression analysis on the motion control command and the fundamental energy of the first signal component within the same time window to construct a mapping function; S43: Calculate the fundamental energy change expected by the current motion control command based on the mapping function, and identify the signal component corresponding to the change as the load fluctuation component.

[0058] In step S41, the motion control command can be the torque output value or speed setpoint of the joint motor. For example, when the robot performs actions such as grasping, moving, or placing, its joint motors will generate corresponding torque output or speed changes. These values ​​can directly quantify the robot's load state, and their purpose is to provide input data that is highly correlated with the robot's actions as the basis for establishing a mapping relationship.

[0059] In step S42, multiple regression analysis is a statistical method used to analyze the relationship between multiple independent variables (i.e., motion control commands) and a dependent variable (i.e., the fundamental wave energy in the first signal component). Through this analysis, a mathematical model, namely the mapping function, can be constructed, which can quantify the influence of different combinations of motion control commands on the fundamental wave energy. In one specific implementation, when constructing the mapping function between motion control commands and the fundamental energy of the power supply, an experiment can be designed to collect data first. For multiple command input terminals, for example, command A is a torque or speed value given by analog voltage, taking values ​​of 0.1V, 0.5V, and 1.0V; command B is a speed command given by frequency, taking values ​​of 10Hz, 50Hz, and 100Hz; command C is the duration or pulse width of the action, taking values ​​of 1ms, 5ms, and 10ms. Any one of the commands A, B, and C can be randomly selected to form a combination, resulting in multiple command combinations. The corresponding fundamental energy is measured under each combination, and the average value is taken after repeated measurements. Subsequently, the data is cleaned, outliers are removed, and the data is standardized to eliminate dimensional differences. Then, the standardized commands A, B, and C are used as independent variables, and the fundamental energy is used as the dependent variable. A multiple regression model is constructed using the least squares method, such as fundamental energy = β0 + β1·command A + β2·command B + β3·command C + ε. Here, β0 is the baseline expected value of the fundamental energy when all commands are 0. In a real-world physical scenario, it represents the unloaded fundamental energy of the power supply when there are no motion commands; β1 is the regression coefficient of command A; β2 is the regression coefficient of command B; and β3 is the regression coefficient of command C. This multivariate regression model is the required mapping function, which can be used to predict the expected fundamental energy value based on any combination of commands.

[0060] The same time window refers to aligning motion control command data and fundamental wave energy data in time during data acquisition and analysis to ensure that they correspond to the same physical event or time period. The purpose is to establish an accurate and predictive mapping model to accurately describe the dynamic relationship between robot actions and power supply fundamental wave energy.

[0061] In step S43, once the mapping function is constructed, the current real-time motion control command can be used as input to calculate the expected change in the fundamental power supply energy under the current robot action. This expected change is considered to be a signal component entirely caused by the robot action and is therefore identified as a load fluctuation component. The purpose is to accurately separate the interference caused by the robot action from the complex power supply signal, providing a clean input for subsequent background degradation signal extraction.

[0062] Through the above technical solution, this application can significantly improve the identification accuracy and separation effect of load fluctuation components in the operating data of industrial robot power supplies. By employing motion control commands that directly reflect robot actions and combining them with multivariate regression analysis to construct an accurate mapping function, the fundamental energy changes caused by robot actions can be more accurately quantified and predicted. Therefore, load fluctuation components can be more effectively removed from the first signal component, minimizing the interference of robot actions on the internal operating status signal of the power supply. This provides a cleaner data foundation for subsequent extraction of the background degradation signal characterizing the physical properties of components, resulting in more accurate and reliable health assessment results for the power supply system and effectively avoiding erroneous maintenance decisions caused by misjudging load fluctuations as component degradation.

[0063] In step S5, the background degradation signal is a signal that characterizes the degradation of the physical characteristics of the internal components of the power supply after removing the load fluctuation component. It is a key indicator for assessing the health status of the power supply.

[0064] Specifically, step S5 includes: S51: Subtract the time-domain waveform of the load fluctuation component from the first signal component in the time domain to obtain the residual signal; S52: Perform adaptive filtering on the residual signal to eliminate the fundamental frequency fluctuation component remaining due to mapping error and output a smoothed background degradation signal.

[0065] In step S51, a time-domain subtraction operation is performed between the first signal component and the load fluctuation component to initially remove the load fluctuation effect caused by the industrial robot's actions. The first signal component is a signal characterizing the internal operating state of the power supply, while the load fluctuation component is a signal component caused by the industrial robot's actions, identified based on the mapping relationship between motion control commands and fundamental energy. By directly subtracting the time-domain waveform of the load fluctuation component, a residual signal with the load effect initially eliminated can be obtained.

[0066] In step S52, adaptive filtering refers to a filtering method that can dynamically adjust filtering parameters according to signal characteristics, such as using Least Mean Square (LMS). Specifically, the central controller integrates an adaptive filter module, which consists of a digital signal processor (DSP). The DSP runs the LMS algorithm internally. The central controller generates a sine wave signal using an oscillator related to the fundamental frequency of the mapping error. The frequency of this oscillator can be dynamically adjusted according to system calibration results or real-time spectrum analysis to match the fundamental frequency of the mapping error. The residual signal is used as the input to the adaptive filter, and the generated sine wave is used as the reference input. The LMS algorithm iteratively adjusts the filter coefficients to minimize the error between the filter output and the residual signal. This process aims to separate the part of the residual signal that is similar to the frequency component of the reference signal (i.e., the fundamental frequency fluctuations left by the mapping error) from the residual signal. The error signal output by the adaptive filter is the signal with the fundamental frequency fluctuation component removed. This signal is then further smoothed by a low-pass filter to eliminate any other possible high-frequency noise, ultimately resulting in a smooth, degraded signal. The amplitude or phase change of this background degradation signal can be directly used to characterize the degradation of the physical properties of the internal components of the power supply.

[0067] In step S6, the health assessment results are a judgment on the current and future health status of the power system based on the evolution trend of the background degradation signal.

[0068] Specifically, step S6 includes: S61: Divide the background degradation signal into segments according to a fixed time window, and extract the amplitude sequence of multiple characteristic frequencies related to the aging of power supply components within each window; S62: Fit the exponential growth curves of each amplitude sequence and calculate its growth rate; S63: When the growth rate of at least one characteristic frequency exceeds a preset rate threshold, calculate the comprehensive health index by combining the weight coefficients of each characteristic frequency, and output the health level according to the interval in which the comprehensive health index is located.

[0069] In step S61, segmenting the background degradation signal into fixed time windows means dividing the continuously acquired background degradation signal into a series of data segments with the same duration. The purpose is to ensure the consistency of the data analysis benchmark when feature extraction is performed at different time points, thereby facilitating comparison and tracking of signal changes caused by component degradation. Specifically, extracting the amplitude sequence of multiple characteristic frequencies related to the aging of power supply components within each window can be understood as identifying the frequency components unique to the changes in physical characteristics (such as increased ESR of capacitors, inductor saturation, etc.) during the aging process of power supply components through spectrum analysis (e.g., Fast Fourier Transform), and recording the amplitude of these frequency components within each time window. These characteristic frequencies are sensitive indicators of the health status of components, and changes in their amplitude directly reflect the degree of component degradation.

[0070] In step S62, fitting the exponential growth curve for each amplitude sequence involves using mathematical methods (such as least squares) to fit an exponential function curve describing the change of amplitude over time for each characteristic frequency, based on the amplitude data points extracted within different time windows. The purpose is to accurately capture the nonlinear trend of amplitude changes during component aging, as many physical degradation processes exhibit exponential or near-exponential growth patterns. Calculating the growth rate involves determining the slope of the fitted exponential growth curve at a specific time point or time period to quantify the rate at which the amplitude of that characteristic frequency increases over time. The growth rate directly reflects the speed of component degradation.

[0071] In step S63, when the growth rate of at least one characteristic frequency exceeds a preset rate threshold, a critical value is set. Once the amplitude growth rate of a characteristic frequency related to component aging reaches or exceeds this critical value, the degradation rate of the component is considered to have reached a level requiring attention. This preset rate threshold can be empirically set or determined experimentally based on the component type, operating environment, design life, and historical failure data. Calculating the comprehensive health index by combining the weighting coefficients of each characteristic frequency means assigning a weighting coefficient to each characteristic frequency, taking into account the differences in the impact of different components or different degradation modes on the overall health of the power system. For example, higher weights are given to critical components or degradation modes that may lead to serious failures.

[0072] The comprehensive health index is calculated as a weighted sum of the growth rates of various characteristic frequencies and their corresponding weighting coefficients. Its purpose is to provide a single quantitative indicator that comprehensively reflects the overall health status of the power system. The health level is output based on the interval in which the comprehensive health index falls. This involves comparing the calculated comprehensive health index with multiple pre-defined intervals, each interval corresponding to a health level, such as normal, warning, minor fault, and severe fault. This allows for a clear and intuitive display of the current health status of the power system to users or maintenance personnel, providing a basis for subsequent maintenance decisions.

[0073] Through the above technical solution, this application provides a more accurate, objective, and forward-looking method for health assessment of industrial robot power systems. This method, by quantitatively analyzing characteristic frequencies related to component aging in the background degradation signal, can effectively identify and track early signs of component degradation, avoiding the subjectivity and lag in health assessment results of traditional methods. Specifically, by monitoring the amplitude sequence and exponential growth rate of multiple characteristic frequencies, refined management of the degradation process of different types of components can be achieved, thus providing sufficient early warning time before failures occur. Furthermore, by combining weighted coefficients to calculate a comprehensive health index and output a health level, the assessment results become more instructive, helping maintenance personnel to develop more scientific and reasonable maintenance plans, reducing the risk of unplanned downtime, extending the service life of the power system, and ultimately improving the operational reliability and production efficiency of industrial robots.

[0074] Please refer to Figure 2 , Figure 3 This application also provides an industrial robot power supply operation data acquisition and analysis system, which is used to implement the steps of any of the above methods, and the system includes: Acquisition module 201: Synchronously acquires electrical parameter signals of multiple power input nodes and power output nodes within the power system; Time-frequency transformation module 202: Performs time-frequency domain transformation processing on multiple electrical parameter signals to extract the frequency distribution characteristics of the dynamic evolution of each electrical parameter signal on the time axis, and obtains multi-channel time-frequency feature vectors; Separation module 203: Performs independent component decomposition on the multi-channel time-frequency feature vector to separate the first signal component characterizing the internal operating state of the power supply; Mapping relationship establishment module 204: acquires the real-time motion control commands of the industrial robot and establishes a mapping relationship between the motion control commands and the fundamental wave energy in the first signal component, so as to identify the load fluctuation component caused by the robot's actions; Extraction module 205: Removes the load fluctuation component from the first signal component and extracts the background degradation signal that characterizes the physical properties of the components; Evaluation Module 206: Based on the evolution trend of the background degradation signal over time, determine the health assessment results of the power system.

[0075] The industrial robot power supply operation data acquisition and analysis system of this application achieves comprehensive monitoring and diagnosis of the power supply's operating status through modular design. Specifically, the acquisition module 201 is first responsible for high-precision synchronous acquisition of electrical parameter signals from the power supply's input and output terminals, providing basic data for subsequent analysis. Subsequently, the time-frequency conversion module 202 processes these raw signals, converting them into time-frequency domain features, thereby capturing rich dynamic frequency information in the signals. Based on this, the separation module 203 uses independent component decomposition technology to accurately separate the first signal component characterizing the internal operating status of the power supply from the complex mixed signal, effectively filtering out external interference. To eliminate the influence of the robot's own movements on the power supply signal, the mapping relationship establishment module 204 acquires the robot's motion control commands in real time and establishes a mapping relationship between them and the fundamental energy in the first signal component, thereby identifying and quantifying the load fluctuation components caused by the robot's movements. Next, the extraction module 205 removes these load fluctuation components from the first signal component, so that the remaining signal can more purely reflect the physical degradation characteristics of the internal components of the power supply, i.e., the background degradation signal. Finally, the evaluation module 206 conducts an in-depth analysis of the evolution trend of the background degradation signal over time, such as monitoring changes in its amplitude, frequency, and other characteristics, thereby determining the health assessment results of the power system. Through the above collaborative work, this system ensures the accuracy and reliability of the power system health assessment.

[0076] Through the aforementioned system solution, this application provides an integrated and automated industrial robot power supply operation data acquisition and analysis platform. Through the close collaboration of its various functional modules, the system achieves fully automated processing from raw data acquisition to final health assessment, significantly improving the efficiency and accuracy of data analysis. Specifically, the system can effectively distinguish between normal load fluctuations caused by robot movements and actual degradation of internal power supply components, avoiding misjudgments and thus making power supply health assessment results more accurate and reliable. Furthermore, the modular design gives the system good scalability and maintainability, facilitating functional upgrades or adjustments according to actual needs. Therefore, this system provides strong technical support for predictive maintenance and fault diagnosis of industrial robot power supplies, helping to extend equipment life, reduce operating costs, and improve the stability and safety of industrial production.

[0077] The above description is merely an embodiment of this application and is not intended to limit the scope of protection of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A method for collecting and analyzing power operation data of an industrial robot, characterized by, The method includes the following steps: S1: Synchronously acquire electrical parameter signals of multiple power input nodes and power output nodes within the power system; S2: Perform time-frequency domain transformation processing on the multiple electrical parameter signals to extract the frequency distribution characteristics of the dynamic evolution of each electrical parameter signal on the time axis, and obtain a multi-channel time-frequency feature vector; S3: Perform independent component decomposition on the multi-channel time-frequency feature vector to separate the first signal component that characterizes the internal operating state of the power supply. S4: Obtain the real-time motion control command of the industrial robot and establish the mapping relationship between the motion control command and the fundamental energy in the first signal component to identify the load fluctuation component caused by the robot's action. S5: Remove the load fluctuation component from the first signal component and extract the background degradation signal that characterizes the physical properties of the component; S6: Based on the evolution trend of the background degradation signal in the time series, determine the health assessment result of the power system.

2. The industrial robot power operation data collection and analysis method according to claim 1, characterized in that, Step S4 includes: S41: Real-time acquisition of the torque output value or speed setpoint of the joint motor as the motion control command; S42: Perform multivariate regression analysis on the motion control command and the fundamental energy of the first signal component within the same time window to construct a mapping function; S43: Calculate the fundamental energy change expected by the current motion control command according to the mapping function, and identify the signal component corresponding to the change as the load fluctuation component.

3. The method of claim 1, wherein the method further comprises: Step S5 includes: S51: Subtract the time-domain waveform of the load fluctuation component from the first signal component in the time domain to obtain the residual signal; S52: Perform adaptive filtering on the residual signal to eliminate the fundamental frequency fluctuation component remaining due to mapping error, and output a smoothed background degradation signal.

4. The method of claim 1, wherein the method further comprises: Step S6 includes: S61: The background degradation signal is segmented according to a fixed time window, and the amplitude sequence of multiple characteristic frequencies related to the aging of power supply components within each window is extracted; S62: Fit the exponential growth curves of each amplitude sequence and calculate its growth rate; S63: When the growth rate of at least one characteristic frequency exceeds a preset rate threshold, a comprehensive health index is calculated by combining the weighting coefficients of each characteristic frequency, and a health level is output according to the interval in which the comprehensive health index is located.

5. The method of claim 1, wherein the method further comprises: Step S1 includes: S11: The electrical parameter signals of the power input node and the power output node are synchronously captured using a sampling frequency of not less than 100kHz to obtain an original sampling sequence containing the transient voltage spike characteristics of the external power grid and the weak high-frequency ripple characteristics generated by the aging of internal components of the power supply. The original sampling sequence is the digital expression of the electrical parameter signals.

6. The method of claim 1, wherein the method further comprises: Step S1 also includes: S12: The central controller sends a synchronous sampling trigger signal to the analog-to-digital converter of the sensor deployed at the power input node and the power output node to ensure the physical synchronization of the sampling actions of each channel; S13: Associate each acquired data point with a nanosecond-level timestamp, and perform phase compensation on the electrical parameter signals of different channels based on the timestamps to eliminate the differences in signal propagation delay at different points.

7. The method of claim 6, wherein the method further comprises: In step S13, associating a nanosecond-level timestamp with each collected data point includes the following steps: S131: Send a global synchronization pulse to all sensors; S132: Record the local count value of the global synchronization pulse received by each of the sensors, and calculate the clock drift coefficient of each sensor in combination with the standard clock frequency of the central controller; S133: Perform linear compensation on the original sampling time of each of the sensors according to the clock drift coefficient to generate the nanosecond-level timestamp under a unified clock reference.

8. The method of claim 1, wherein the method further comprises: Step S2 includes: S21: Use short-time Fourier transform to perform frame-by-frame windowing processing on each of the electrical parameter signals to obtain the spectral distribution within each time window; S22: Arrange the spectral distribution of all time windows within the same channel in chronological order to form the time-frequency characteristic sequence of that channel; S23: Combine the time-frequency feature sequences of all channels into the multi-channel time-frequency feature vector.

9. The industrial robot power operation data collection and analysis method of claim 1, wherein, Step S3 includes: S31: Extract multiple statistically independent source signal vectors from the multi-channel time-frequency feature vector using the independent component analysis algorithm; S32: Calculate the energy attenuation ratio of each source signal vector between the power input node and the power output node, and determine the source signal vector with an energy attenuation ratio greater than a preset threshold and whose time-frequency characteristics are instantaneous wideband pulses as the first signal component.

10. An industrial robot power supply operation data collection and analysis system, characterized by The system is used to implement the steps of the method according to any one of claims 1-9, and the system includes: Acquisition module: Synchronously acquires electrical parameter signals from multiple power input nodes and power output nodes within the power system; Time-frequency transformation module: Performs time-frequency domain transformation processing on the multiple electrical parameter signals to extract the frequency distribution characteristics of the dynamic evolution of each electrical parameter signal on the time axis, and obtains a multi-channel time-frequency feature vector; Separation module: Performs independent component decomposition on the multi-channel time-frequency feature vector to separate the first signal component characterizing the internal operating state of the power supply; Mapping relationship establishment module: acquires the real-time motion control commands of the industrial robot and establishes a mapping relationship between the motion control commands and the fundamental wave energy in the first signal component, so as to identify the load fluctuation component caused by the robot's actions; Extraction module: Removes the load fluctuation component from the first signal component and extracts the background degradation signal characterizing the physical properties of the component; Evaluation module: Based on the evolution trend of the background degradation signal over time, determine the health assessment result of the power system.

Images