A method and system for diagnosing faults in industrial robot reducers

Abstract

This invention provides a method and system for fault diagnosis of industrial robot reducers. The method includes acquiring vibration signals and determining an initial population using a preset whale optimization algorithm; performing a first VMD decomposition on the vibration signals based on the VMD parameters of individuals within the initial population, and calculating a first fitness by combining Hilbert transformation to obtain signal envelope correlation features; screening the initial population based on the first fitness to determine a potential solution population; performing a second VMD decomposition on the vibration signals based on the VMD parameters in the potential solution population and calculating fitness using fault impact correlation feature indicators; further optimizing the potential solution population using the preset whale optimization algorithm to obtain globally optimal VMD parameters; and determining fault information based on the globally optimal VMD parameters, the vibration signals, and the theoretical characteristics of the robot reducer. This invention solves the problem of existing industrial robot reducer fault diagnosis methods lacking strong global search capabilities, objective feature extraction, and high computational efficiency.

Description

Technical Field

[0001] This invention relates to the field of robot condition monitoring and intelligent operation and maintenance technology, and in particular to a method and system for diagnosing faults in industrial robot reducers. Background Technology

[0002] Industrial robots are core equipment in high-precision manufacturing, and the reducers at their joints are key components ensuring motion accuracy and load-bearing capacity. Reducers operate under heavy and variable load conditions for extended periods, making their internal gears, bearings, and other components prone to pitting, peeling, cracking, and tooth breakage. These faults initially manifest as weak impact signals, easily masked by strong background noise, vibrations from other components, and complex transmission paths.

[0003] Traditional signal processing methods (such as wavelet transform and empirical mode decomposition) suffer from problems such as mode aliasing and endpoint effects when processing non-stationary and nonlinear signals, resulting in insufficient fault feature extraction. Although variational mode decomposition (VMD) can effectively overcome the shortcomings of empirical mode decomposition, its performance is highly dependent on the preset number of modes K and the penalty parameter α.

[0004] Existing technologies employ intelligent algorithms (such as Particle Swarm Optimization and Beluga Optimization) to optimize VMD parameters. However, these methods suffer from several drawbacks: First, they are prone to premature convergence. For example, the BWO algorithm relies on fixed probability thresholds for stage switching, making it susceptible to getting trapped in local optima and resulting in insufficient parameter optimization. Second, feature selection is subjective. Using methods like TOPSIS to select components requires manually setting indicator weights, leading to poor adaptability. Finally, computational efficiency is a significant bottleneck, representing a core problem that has not been fully addressed in existing technologies. VMD computation is complex, its core being the solution of a constrained variational problem through ADMM iteration, resulting in high computational costs. Optimization algorithms require repeated calls to VMD, often thousands of times, leading to excessively long total computation time and failing to meet the real-time requirements of industrial online monitoring. Simply reducing the population size or the number of iterations directly sacrifices diagnostic accuracy, creating a trade-off between accuracy and efficiency. Summary of the Invention

[0005] Based on this, the purpose of this invention is to provide a method and system for diagnosing faults in industrial robot reducers, aiming to solve the problem that existing technologies lack methods for diagnosing faults in industrial robot reducers with strong global search capabilities, objective feature extraction, and high computational efficiency.

[0006] A method for diagnosing faults in an industrial robot reducer according to an embodiment of the present invention, the method comprising:

[0007] The vibration signal of the industrial robot reducer in motion is acquired by a preset sensor, and an initial population is determined based on preset parameters using a preset whale optimization algorithm. The position of each individual in the initial population represents a set of VMD parameters to be optimized.

[0008] Based on the VMD parameters of individuals in the initial population, the vibration signal is decomposed into a first VMD, and the signal envelope correlation features are obtained by combining Hilbert transformation to calculate the first fitness. The initial population is then screened based on the first fitness to determine potential solution populations.

[0009] Based on the VMD parameters corresponding to individuals in the potential solution population, the vibration signal is decomposed into a second VMD, and the fitness is calculated using fault impact related characteristic indicators. Then, the potential solution population is optimized by combining the population update mechanism of the preset whale optimization algorithm and the global optimal individual perturbation strategy to obtain the global optimal VMD parameters.

[0010] The vibration signal is decomposed into a second VMD based on the globally optimal VMD parameters to obtain the intrinsic mode functions. Sensitive intrinsic multimodal function components are then selected by multi-scale permutation entropy to determine fault information based on the sensitive intrinsic multimodal function components and the theoretical characteristics of the robot reducer.

[0011] In addition, the industrial robot reducer fault diagnosis method according to the above embodiments of the present invention may also have the following additional technical features:

[0012] Furthermore, based on the VMD parameters of individuals in the initialized population, the vibration signal is subjected to a first VMD decomposition, and the signal envelope correlation features are obtained by combining Hilbert transformation to calculate a first fitness. The step of screening the initialized population to determine potential solution populations based on the first fitness includes:

[0013] The vibration signal is segmented to determine the feature signal segment, and then the feature signal segment is decomposed into the first VMD based on the VMD parameters of the individuals in the initial population to obtain the first intrinsic multimodal function component. The convergence accuracy of the ADMM iteration inside the VMD is reduced when the first VMD decomposition is performed.

[0014] The envelope signal and corresponding envelope spectrum are determined by performing a Hilbert transform on the first intrinsic multimodal function component;

[0015] The first fitness is calculated based on the envelope spectrum, and the initial population is sorted from largest to smallest according to the value of the first fitness. Individuals within the top preset proportion of fitness ranking are selected to form a potential solution population.

[0016] The formulas for calculating the envelope signal and envelope spectrum are as follows:

[0017]

[0018]

[0019] in, For envelope signal, For the first intrinsic multimodal function component, For envelope spectrum, H For Hilbert transform, This is a Fourier transform.

[0020] Furthermore, the formula for calculating the first fitness is:

[0021]

[0022] in, The theoretical fault characteristic frequency, This refers to the domain of theoretical fault characteristic frequencies. For individuals The VMD mode decomposition number specified in the code. For the first k The envelope spectrum of the IMF component in the th f The amplitude at each frequency point.

[0023] Furthermore, based on the VMD parameters corresponding to individuals in the potential solution population, a second VMD decomposition is performed on the vibration signal, and the fitness is calculated using fault impact-related characteristic indicators. Then, the potential solution population is optimized by combining the population update mechanism of the preset whale optimization algorithm and the globally optimal individual perturbation strategy to obtain the globally optimal VMD parameters. The steps include:

[0024] The vibration signal is decomposed into a second intrinsic multimodal function component based on the VMD parameters of individuals in the potential solution population, and the convergence accuracy of the ADMM iteration inside the VMD is restored when performing the second VMD decomposition.

[0025] Based on the vibration signal, a corresponding fitness function is selected to perform a second fitness on the second intrinsic multimodal function component;

[0026] The first current optimal individual is determined based on the second fitness, and the corresponding population update mechanism is selected based on the probability and coefficient vector of the preset whale optimization algorithm. The position of individuals in the potential solution population is updated based on the first current optimal individual and the corresponding population update mechanism to determine the intermediate solution population, and the updated second current optimal individual is determined.

[0027] A perturbation is applied to the second current best individual to determine a third current best individual, and the individual with higher fitness between the second and third current best individuals is determined as the fourth current best individual, so that the fourth current best individual replaces the second current best individual;

[0028] The fourth current best individual and the intermediate solution population constitute a new potential solution population, so as to return to the step of performing the second VMD decomposition on the vibration signal based on the VMD parameters of the individuals in the potential solution population to obtain the second intrinsic multimodal function components, until the number of iterations reaches the preset iteration threshold, and output the globally optimal VMD parameters.

[0029] Furthermore, the step of performing a second VMD decomposition on the vibration signal based on the globally optimal VMD parameters to obtain the intrinsic mode functions, and then filtering out the sensitive intrinsic multimode function components through multi-scale permutation entropy, includes:

[0030] The vibration signal is decomposed into intrinsic mode functions and corresponding third intrinsic multimode function components by performing a second VMD decomposition based on the globally optimal VMD parameters, and the convergence accuracy of the internal ADMM iteration of VMD is restored when performing the second VMD decomposition.

[0031] Calculate the coarse-grained sequence of each of the third intrinsic multimodal function components under a preset scaling factor;

[0032] After reconstructing the phase space of each of the coarse-grained sequences, the corresponding permutation entropy is calculated.

[0033] The third intrinsic multimodal function component with the smallest entropy value at the key scale is selected as the sensitive intrinsic multimodal function component.

[0034] Furthermore, the step of determining fault information based on the sensitive intrinsic multimodal function components and the theoretical characteristics of the robot reducer includes:

[0035] Hilbert envelope demodulation is performed on the sensitive intrinsic multimodal function components to determine the target envelope spectrum;

[0036] The significant spectral peak frequencies in the target envelope spectrum are matched with the theoretical fault characteristic frequency set;

[0037] If the significant spectral peak frequency matches any type of theoretical fault characteristic frequency and its harmonics, then the corresponding fault is determined to have occurred.

[0038] Furthermore, the steps for determining the initial population based on preset parameters using a preset whale optimization algorithm include:

[0039] Define the population size N and the maximum number of iterations T; define the parameter search space: , The population was initialized using Sobol sequences.

[0040] Another object of the present invention is an industrial robot reducer fault diagnosis system, the system comprising:

[0041] The data acquisition module is used to acquire the vibration signal of the industrial robot reducer in motion state collected by the preset sensor, and determine the initial population based on the preset parameters and the preset whale optimization algorithm. The position of each individual in the initial population represents a set of VMD parameters to be optimized.

[0042] The data filtering module is used to perform a first VMD decomposition on the vibration signal based on the VMD parameters of individuals in the initial population, and to obtain signal envelope correlation features by combining Hilbert transformation to calculate a first fitness, and to filter the initial population to determine a potential solution population based on the first fitness.

[0043] The data optimization module is used to perform a second VMD decomposition on the vibration signal based on the VMD parameters corresponding to individuals in the potential solution population, calculate the fitness using fault impact related characteristic indicators, and then optimize the potential solution population by combining the population update mechanism of the preset whale optimization algorithm and the global optimal individual perturbation strategy to obtain the global optimal VMD parameters.

[0044] The fault diagnosis module is used to perform a second VMD decomposition on the vibration signal based on the globally optimal VMD parameters to obtain the intrinsic mode functions, and to filter out sensitive intrinsic multimodal function components through multi-scale permutation entropy, so as to determine fault information based on the sensitive intrinsic multimodal function components and the theoretical characteristics of the robot reducer.

[0045] Another objective of this invention is to provide a storage medium storing a computer program that, when executed by a processor, implements the steps of the above-described industrial robot reducer fault diagnosis method.

[0046] Another objective of this invention is to provide an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of the above-described industrial robot reducer fault diagnosis method.

[0047] This invention employs a two-stage optimization strategy. In the early stage, a low-cost pre-screening mechanism significantly reduces the number of high-cost, high-precision VMD decompositions, fundamentally solving the computational efficiency bottleneck when optimizing VMD parameters. The subsequent high-precision VMD decomposition ensures data accuracy, and Sobol sequence initialization guarantees the uniformity of the initial population distribution, laying the foundation for global search. Furthermore, a perturbation strategy provides a powerful local extremum escape mechanism. The combination of these two approaches ensures the algorithm possesses both excellent global exploration and local exploitation capabilities, effectively avoiding premature convergence. In addition, by employing multi-scale permutation entropy to screen sensitive intrinsic mode function components, this screening process is entirely based on the dynamic complexity of the vibration signal itself, eliminating the need for manually setting index weights. It can adaptively capture signal dynamic abrupt changes caused by reducer faults, accurately identifying the sensitive components richest in fault information. Even in scenarios where the original vibration signal is submerged in strong noise, it can still stably extract fault features, exhibiting significantly better anti-interference capabilities than traditional feature screening methods. Therefore, this invention solves the problem of the lack of existing industrial robot reducer fault diagnosis methods with strong global search capabilities, objective feature extraction, and high computational efficiency. Attached Figure Description

[0048] Figure 1 This is a flowchart of the industrial robot reducer fault diagnosis method in the first embodiment of the present invention;

[0049] Figure 2 This is a schematic diagram of the results of the industrial robot reducer fault diagnosis system in the second embodiment of the present invention;

[0050] Figure 3 This is a schematic diagram of the structure of the electronic device in the third embodiment of the present invention;

[0051] The following detailed description, in conjunction with the accompanying drawings, will further illustrate the present invention. Detailed Implementation

[0052] To facilitate understanding of the present invention, a more complete description will be given below with reference to the accompanying drawings. Several embodiments of the invention are illustrated in the drawings. However, the invention can be implemented in many different forms and is not limited to the embodiments described herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete.

[0053] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used herein in the description of the invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0054] Example 1

[0055] Please see Figure 1 The figure shows a fault diagnosis method for an industrial robot reducer in the first embodiment of the present invention, which specifically includes steps S01-S04.

[0056] S01, acquire the vibration signal of the industrial robot reducer in motion state collected by the preset sensor, and determine the initial population based on the preset parameters and the preset whale optimization algorithm. The position of each individual in the initial population represents a set of VMD parameters to be optimized.

[0057] Specifically, set the population size N0 and the maximum number of iterations T; define the parameter search space: , The population was initialized using Sobol sequences (a type of deterministic, low-difference sequence), and the location of each individual was determined. This represents a set of VMD parameters to be optimized. In practice, accelerometers can be installed on the joint housings of the industrial robot to collect data and vibration signals during operation. Furthermore, by uniformly distributing individuals in the parameter space using a Sobol sequence, more potential optimal solution regions are covered, preventing the algorithm from getting stuck in local searches in the initial stage and improving the global optimization capability in subsequent optimizations.

[0058] S02, based on the VMD parameters of individuals in the initial population, the vibration signal is decomposed into a first VMD, and the signal envelope correlation features are obtained by combining Hilbert transformation to calculate the first fitness. The initial population is then screened according to the first fitness to determine the potential solution population.

[0059] Specifically, the vibration signal is segmented to determine feature signal segments. Then, based on the VMD parameters of individuals in the initial population, the feature signal segments are decomposed using the first VMD to obtain first intrinsic multimodal function components. During the first VMD decomposition, the convergence accuracy of the ADMM iteration within the VMD is reduced. The first intrinsic multimodal function components are subjected to Hilbert transform to determine the envelope signal and the corresponding envelope spectrum. The first fitness is calculated based on the envelope spectrum, and the initial population is sorted from largest to smallest according to the value of the first fitness. Individuals within the top preset proportion of fitness ranking are selected to form a potential solution population.

[0060] The formulas for calculating the envelope signal and envelope spectrum are as follows:

[0061]

[0062]

[0063] in, For envelope signal, For the first intrinsic multimodal function component, For envelope spectrum, H For Hilbert transform, This is a Fourier transform.

[0064] The formula for calculating the first fitness is:

[0065]

[0066] in, The theoretical fault characteristic frequency, This refers to the domain of theoretical fault characteristic frequencies. For individuals The VMD mode decomposition number specified in the code. For the first k The envelope spectrum of the IMF component in the th f The amplitude at each frequency point.

[0067] In practical implementation, the vibration signal is first downsampled to use shorter, representative signal segments to represent the whole for subsequent calculations, thereby reducing the amount of data processing. Furthermore, during VMD decomposition, the convergence accuracy of the ADMM iteration within VMD is reduced, thus quickly obtaining the first intrinsic multimodal function components. Then, the envelope spectra of all first intrinsic multimodal function components are calculated at known theoretical fault characteristic frequencies. Neighborhood The sum of amplitude values ​​within a given frequency range is used as the first fitness. The purpose of selecting the sum of amplitude values ​​as the first fitness is to avoid complex statistical calculations; it only requires "amplitude accumulation" of the approximate envelope spectrum within a specified frequency range. The physical meaning of this fitness is the "degree of energy concentration near the fault characteristic frequency," directly targeting the core requirement of "whether the parameters can be decomposed into components containing fault information," thus avoiding interference from irrelevant indicators. While the VMD decomposition at this stage is a "simplified version" with limited data precision, the "sum of amplitude values" only needs to determine the "energy concentration trend," not a precise numerical value. Even with some error in the envelope spectrum, the trend of "higher sum of amplitude values ​​near the fault frequency" can still be stably reflected, preventing the screening from failing due to data simplification. Furthermore, through appropriate first fitness selection and population ranking based on the first fitness, only the top 50%~70% of "potential solutions" are retained for the next generation, while the rest are directly eliminated. This mechanism significantly reduces the number of individuals requiring precise calculations in the early stages.

[0068] S03, based on the VMD parameters corresponding to individuals in the potential solution population, the vibration signal is decomposed into a second VMD, and the fitness is calculated using fault impact related characteristic indicators. Then, the potential solution population is optimized by combining the population update mechanism of the preset whale optimization algorithm and the global optimal individual perturbation strategy to obtain the global optimal VMD parameters.

[0069] Specifically, based on the VMD parameters of individuals in the potential solution population, the vibration signal is decomposed using the second VMD method to obtain the second intrinsic multimodal function components, and the convergence accuracy of the ADMM iteration within the VMD is restored during the second VMD decomposition. A corresponding fitness function is selected based on the vibration signal to apply a second fitness to the second intrinsic multimodal function components. A first current optimal individual is determined based on the second fitness, and a corresponding population update mechanism is selected based on the probability and coefficient vector magnitude of the preset whale optimization algorithm. The position of individuals in the potential solution population is updated and determined based on the first current optimal individual and the corresponding population update mechanism. An intermediate solution population is established, and an updated second current best individual is determined. A perturbation is applied to the second current best individual to determine a third current best individual. The individual with the higher fitness between the second and third current best individuals is then designated as the fourth current best individual, replacing the second current best individual. The fourth current best individual and the intermediate solution population constitute a new potential solution population. The process then returns to the step of performing the second VMD decomposition on the vibration signal based on the VMD parameters of individuals in the potential solution population to obtain the second intrinsic multimodal function components, until the number of iterations reaches a preset iteration threshold, and the globally optimal VMD parameters are output.

[0070] In practical implementation, a second VMD decomposition is performed on the complete vibration signal based on the VMD parameters of individuals within the potential solution population to ensure the accuracy of the obtained second intrinsic multimodal function components. Then, a suitable fitness function is determined based on the characteristics of the vibration signal for second fitness calculation to better reflect fault characteristics and ensure optimization accuracy. Subsequently, the population is iterated using a pre-defined whale optimization algorithm to find the globally optimal VMD parameters with high precision within a narrowed search range. A perturbation is applied to the optimal solution after each iteration to help the algorithm escape local optima, avoid premature convergence, and ensure that the parameter optimization accuracy is superior to traditional PSO and BWO algorithms. After multiple iterations, the desired optimal VMD parameters can be obtained.

[0071] Furthermore, in practical implementation, the corresponding fitness function for calculating the second fitness can be:

[0072]

[0073] in, , It is the envelope signal of the component with the smallest envelope entropy among all IMF (Intrinsic Multimodal Function) components. Minimize Equivalent to maximizing , Let be the envelope entropy.

[0074] It can also be:

[0075]

[0076] Where ES is the envelope spectrum of the sensitive IMF (intrinsic multimodal function). N This represents the total length of the envelope spectrum.

[0077] These two fitness functions employ minimizing envelope entropy and maximizing envelope kurtosis as the second fitness function, respectively. Minimizing envelope entropy is suitable when fault characteristics are masked by moderate-intensity noise (not completely submerged), in which case "regular differences" (entropy changes) are more stable than "impact intensity differences" (kurtosis); for fault types of "periodic weak impacts," the impact amplitude is small but the period is stable, and envelope entropy can more sensitively capture this regular change. Maximizing envelope kurtosis is suitable when fault characteristics are "strong impacts" (such as bearing roller breakage or gear tooth breakage), the impact amplitude is large and sudden, and kurtosis can directly amplify this impact difference.

[0078] With lower signal noise and less obscuring of impact characteristics, the peak sensitivity of kurtosis is better.

[0079] In addition, population renewal mechanisms include at least:

[0080] Surrounding prey: After spotting prey, whales will move towards its location, according to the formula:

[0081]

[0082] in, Let be the location of a whale individual in generation t. Each location represents a set of VMD parameters. Let be the position of the globally optimal individual in generation t. and This is the coefficient vector.

[0083] Bubble web attack: Whales spirally exhale bubbles downstream of their prey, forming a bubble web, and then spiral upwards to approach the prey. The corresponding formula is:

[0084]

[0085] in, This represents the distance between the current individual and the globally optimal individual.b It is a spiral constant. l In order to be in A number randomly generated within the range. The equation is a spiral.

[0086] Random search: When no prey is found, whales will wander randomly within their search space, as shown in the formula:

[0087]

[0088] in, This refers to the location of a randomly selected individual whale from the current population. and This is the coefficient vector.

[0089] In addition, the prey encirclement mechanism is used for global search and narrowing the encirclement range, the bubble net attack mechanism is used for local exploration and fine-grained search for the optimal solution, and the random search mechanism is used to avoid local optima and expand the search range.

[0090] Furthermore, the corresponding formula for applying a perturbation to the current global optimal solution after each iteration is:

[0091]

[0092] in, As the second best individual at present, The third best individual at present. Let be a random perturbation term that follows a standard Cauchy distribution, where the location parameter is 0 and the scale parameter is σ. It follows the standard Cauchy distribution and utilizes its long-tail characteristic to generate a large perturbation with a certain probability, which significantly enhances the algorithm's ability to escape local optima.

[0093] S04, perform a second VMD decomposition on the vibration signal according to the globally optimal VMD parameters to obtain the intrinsic mode functions, and filter out the sensitive intrinsic multimodal function components through multi-scale permutation entropy, so as to determine the fault information based on the sensitive intrinsic multimodal function components and the theoretical characteristics of the robot reducer.

[0094] Specifically, based on the globally optimal VMD parameters, the vibration signal is decomposed into intrinsic mode functions and corresponding third intrinsic multimode function components using a second VMD decomposition, while restoring the convergence accuracy of the internal ADMM iteration of the VMD during the second VMD decomposition; coarse-grained sequences of each of the third intrinsic multimode function components are calculated under a preset scale factor; phase space reconstruction is performed on each of the coarse-grained sequences, and the corresponding permutation entropy is calculated; the third intrinsic multimode function component with the smallest permutation entropy value at the key scale is selected as the sensitive intrinsic multimode function component; Hilbert envelope demodulation is performed on the sensitive intrinsic multimode function component to determine the target envelope spectrum; the significant spectral peak frequencies in the target envelope spectrum are matched with the theoretical fault characteristic frequency set; if the significant spectral peak frequencies match any type of theoretical fault characteristic frequency and its harmonics, then the corresponding fault is determined to have occurred.

[0095] In practical implementation, the coarsened sequence of each third intrinsic multimodal function component under the scale factor s is calculated:

[0096]

[0097] in, Let be the magnitude of the k-th intrinsic multimodal function (IMF) at the i-th discrete sampling point, and s be the scale factor. Let be the value of the j-th data point in the coarsened new time series under a scale factor s. Original IMF components Total length (total number of sampling points). This represents the total length of the new sequence after coarsening.

[0098] Furthermore, after obtaining the coarse-grained sequences, each coarse-grained sequence... After reconstructing the phase space, its permutation entropy is calculated. Then, each IMF component is obtained at different scales. s The set of PE values ​​on, where, For the embedding dimension, The delay time is used. The IMF component with the smallest MPE value at the critical scale (usually a lower scale reflecting the main dynamic characteristics of the signal) is selected as the sensitive component. The smallest MPE value indicates that the time series of this component is the most regular and ordered, and most likely to contain periodic fault impacts. The required sensitive intrinsic multimodal function components are then accurately obtained. Hilbert envelope demodulation is then performed on the selected sensitive IMF components to obtain their high-precision envelope spectrum. The significant peak frequencies in the envelope spectrum are analyzed and compared with theoretically calculated bearing fault characteristic frequencies (such as the inner ring, outer ring, and rolling element passage frequencies). If the peak frequency matches a certain type of fault characteristic frequency and its harmonics, the corresponding fault is determined to have occurred. This allows for accurate and rapid diagnosis of industrial robot reducer faults.

[0099] In summary, the industrial robot reducer fault diagnosis method in the above embodiments of the present invention, through a two-stage optimization strategy, significantly reduces the number of high-cost, high-precision VMD decompositions by employing a low-cost pre-screening mechanism in the early stage, thereby fundamentally solving the computational efficiency bottleneck when optimizing VMD parameters. The high-precision VMD decomposition in the later stage ensures data accuracy, and Sobol sequence initialization ensures the uniformity of the initial population distribution, laying the foundation for global search. Furthermore, the perturbation strategy provides a powerful local extremum escape mechanism. The combination of these two aspects ensures that the algorithm possesses both excellent global exploration and local exploitation capabilities, effectively avoiding premature convergence. In addition, by employing multi-scale permutation entropy to screen sensitive intrinsic mode function components, this screening process is entirely based on the dynamic complexity of the vibration signal itself, without the need for manually setting index weights. It can adaptively capture signal dynamic abrupt changes caused by reducer faults, accurately identify the sensitive components richest in fault information, and can still stably extract fault features even in scenarios where the original vibration signal is submerged in strong noise, demonstrating significantly better anti-interference capabilities than traditional feature screening methods. Therefore, this invention solves the problem of the lack of a fault diagnosis method for industrial robot reducers in the prior art that has strong global search capabilities, objective feature extraction, and high computational efficiency.

[0100] Example 2

[0101] Please see Figure 2 The diagram shown is a structural block diagram of the industrial robot reducer fault diagnosis system proposed in the second embodiment of the present invention. The industrial robot reducer fault diagnosis system 200 includes: a data acquisition module 21, a data filtering module 22, a data optimization module 23, and a fault judgment module 24, wherein:

[0102] The data acquisition module 21 is used to acquire the vibration signal of the industrial robot reducer in motion state collected by the preset sensor, and determine the initial population based on the preset parameters and the preset whale optimization algorithm. The position of each individual in the initial population represents a set of VMD parameters to be optimized.

[0103] The data filtering module 22 is used to perform a first VMD decomposition on the vibration signal based on the VMD parameters of individuals in the initial population, and to obtain signal envelope correlation features by combining Hilbert transformation to calculate a first fitness, and to filter the initial population to determine a potential solution population based on the first fitness.

[0104] Data optimization module 23 is used to perform a second VMD decomposition on the vibration signal based on the VMD parameters corresponding to individuals in the potential solution population, calculate the fitness using fault impact related characteristic indicators, and then optimize the potential solution population by combining the population update mechanism of the preset whale optimization algorithm and the global optimal individual perturbation strategy to obtain the global optimal VMD parameters.

[0105] The fault judgment module 24 is used to perform a second VMD decomposition on the vibration signal according to the globally optimal VMD parameters to obtain the intrinsic mode functions, and to filter out the sensitive intrinsic multimodal function components through multi-scale permutation entropy, so as to determine the fault information based on the sensitive intrinsic multimodal function components and the theoretical characteristics of the robot reducer.

[0106] Example 3

[0107] In another aspect, the present invention also proposes an electronic device, please refer to [link to relevant documentation]. Figure 3 The diagram shows an electronic device according to the third embodiment of the present invention, including a memory 20, a processor 10, and a computer program 30 stored in the memory and executable on the processor. When the processor 10 executes the computer program 30, it implements the industrial robot reducer fault diagnosis method as described above.

[0108] In some embodiments, the processor 10 may be a central processing unit (CPU), controller, microcontroller, microprocessor or other data processing chip, used to run program code stored in memory 20 or process data, such as executing access restriction programs.

[0109] The memory 20 includes at least one type of readable storage medium, such as flash memory, hard disk, multimedia card, card-type memory (e.g., SD or DX memory), magnetic memory, magnetic disk, optical disk, etc. In some embodiments, the memory 20 can be an internal storage unit of an electronic device, such as the hard disk of the electronic device. In other embodiments, the memory 20 can also be an external storage device of the electronic device, such as a plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, etc. Furthermore, the memory 20 can include both internal and external storage units of the electronic device. The memory 20 can be used not only to store application software and various types of data of the electronic device, but also to temporarily store data that has been output or will be output.

[0110] It should be pointed out that, Figure 3 The structure shown does not constitute a limitation on the electronic device. In other embodiments, the electronic device may include fewer or more components than shown, or combine certain components, or have different component arrangements.

[0111] This invention also proposes a computer-readable storage medium storing a computer program that, when executed by a processor, implements the industrial robot reducer fault diagnosis method described above.

[0112] Those skilled in the art will understand that the logic and / or steps represented in the flowcharts or otherwise described herein, for example, can be considered as a ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can mean any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0113] More specific examples of computer-readable media (a non-exhaustive list) include: electrical connections (electronic devices) having one or more wires, portable computer disk drives (magnetic devices), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, computer-readable media can even be paper or other suitable media on which the program can be printed, because the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in computer memory.

[0114] It should be understood that various parts of the present invention can be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0115] In the description of this specification, references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0116] The above embodiments merely illustrate several implementation methods of the present invention, and their descriptions are relatively specific and detailed, but they should not be construed as limiting the scope of the present invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these all fall within the protection scope of the present invention. Therefore, the protection scope of this patent should be determined by the appended claims.

Claims

1. A method for diagnosing faults in an industrial robot reducer, characterized in that, The method includes: The vibration signal of the industrial robot reducer in motion is acquired by a preset sensor, and an initial population is determined based on preset parameters using a preset whale optimization algorithm. The position of each individual in the initial population represents a set of VMD parameters to be optimized. Based on the VMD parameters of individuals in the initial population, the vibration signal is decomposed into a first VMD, and the signal envelope correlation features are obtained by combining Hilbert transformation to calculate the first fitness. The initial population is then screened based on the first fitness to determine potential solution populations. Based on the VMD parameters corresponding to individuals in the potential solution population, the vibration signal is decomposed into a second VMD, and the fitness is calculated using fault impact related characteristic indicators. Then, the potential solution population is optimized by combining the population update mechanism of the preset whale optimization algorithm and the global optimal individual perturbation strategy to obtain the global optimal VMD parameters. The vibration signal is decomposed into a second VMD based on the global optimal VMD parameters to obtain the intrinsic mode functions. Sensitive intrinsic multimodal function components are then selected by multi-scale permutation entropy to determine fault information based on the sensitive intrinsic multimodal function components and the theoretical characteristics of the robot reducer. The steps of performing a first VMD decomposition on the vibration signal based on the VMD parameters of individuals in the initialized population, obtaining signal envelope correlation features by combining Hilbert transformation to calculate a first fitness, and screening the initialized population to determine potential solution populations based on the first fitness include: The vibration signal is segmented to determine the feature signal segment, and then the feature signal segment is decomposed into the first VMD based on the VMD parameters of the individuals in the initial population to obtain the first intrinsic multimodal function component. The convergence accuracy of the ADMM iteration inside the VMD is reduced when the first VMD decomposition is performed. The envelope signal and corresponding envelope spectrum are determined by performing a Hilbert transform on the first intrinsic multimodal function component; The first fitness is calculated based on the envelope spectrum, and the initial population is sorted from largest to smallest according to the value of the first fitness. Individuals within the top preset proportion of fitness ranking are selected to form a potential solution population. The formulas for calculating the envelope signal and envelope spectrum are as follows: in, For envelope signal, For the first intrinsic multimodal function component, For envelope spectrum, H For Hilbert transform, Fourier transform; The steps of performing a second VMD decomposition on the vibration signal based on the globally optimal VMD parameters to obtain the intrinsic mode functions, and then filtering out the sensitive intrinsic multimode function components by multi-scale permutation entropy, include: The vibration signal is decomposed into intrinsic mode functions and corresponding third intrinsic multimode function components by performing a second VMD decomposition based on the globally optimal VMD parameters, and the convergence accuracy of the internal ADMM iteration of VMD is restored when performing the second VMD decomposition. Calculate the coarse-grained sequence of each of the third intrinsic multimodal function components under a preset scaling factor; After reconstructing the phase space of each of the coarse-grained sequences, the corresponding permutation entropy is calculated. The third intrinsic multimodal function component with the smallest entropy value at the key scale is selected as the sensitive intrinsic multimodal function component.

2. The industrial robot reducer fault diagnosis method according to claim 1, characterized in that, The formula for calculating the first fitness is: in, The theoretical fault characteristic frequency, This refers to the domain of theoretical fault characteristic frequencies. For individuals The VMD mode decomposition number specified in the code. For the first k The envelope spectrum of the IMF component in the th f The amplitude at each frequency point.

3. The industrial robot reducer fault diagnosis method according to claim 1, characterized in that, Based on the VMD parameters corresponding to individuals in the potential solution population, the vibration signal is subjected to a second VMD decomposition, and the fitness is calculated using fault impact-related characteristic indicators. Then, the potential solution population is optimized by combining the population update mechanism of the preset whale optimization algorithm and the globally optimal individual perturbation strategy to obtain the globally optimal VMD parameters. The steps include: The vibration signal is decomposed into a second intrinsic multimodal function component based on the VMD parameters of individuals in the potential solution population, and the convergence accuracy of the ADMM iteration inside the VMD is restored when performing the second VMD decomposition. Based on the vibration signal, a corresponding fitness function is selected to apply a second fitness to the second intrinsic multimodal function component; The first current optimal individual is determined based on the second fitness, and the corresponding population update mechanism is selected based on the probability and coefficient vector of the preset whale optimization algorithm. The position of individuals in the potential solution population is updated based on the first current optimal individual and the corresponding population update mechanism to determine the intermediate solution population, and the updated second current optimal individual is determined. A perturbation is applied to the second current best individual to determine a third current best individual, and the individual with higher fitness between the second and third current best individuals is determined as the fourth current best individual, so that the fourth current best individual replaces the second current best individual; The fourth current best individual and the intermediate solution population constitute a new potential solution population, so as to return to the step of performing the second VMD decomposition on the vibration signal based on the VMD parameters of the individuals in the potential solution population to obtain the second intrinsic multimodal function components, until the number of iterations reaches the preset iteration threshold, and output the globally optimal VMD parameters.

4. The industrial robot reducer fault diagnosis method according to claim 1, characterized in that, The steps for determining fault information based on the sensitive intrinsic multimodal function components and the theoretical characteristics of the robot reducer include: Hilbert envelope demodulation is performed on the sensitive intrinsic multimodal function components to determine the target envelope spectrum; The significant spectral peak frequencies in the target envelope spectrum are matched with the theoretical fault characteristic frequency set; If the significant spectral peak frequency matches any type of theoretical fault characteristic frequency and its harmonics, then the corresponding fault is determined to have occurred.

5. The industrial robot reducer fault diagnosis method according to claim 2, characterized in that, The steps for determining the initial population based on preset parameters and a preset whale optimization algorithm include: Define the population size N and the maximum number of iterations T; define the parameter search space: , The population was initialized using Sobol sequences.

6. A fault diagnosis system for an industrial robot reducer, characterized in that, The system for implementing the industrial robot reducer fault diagnosis method as described in any one of claims 1 to 5 includes: The data acquisition module is used to acquire the vibration signal of the industrial robot reducer in motion state collected by the preset sensor, and determine the initial population based on the preset parameters and the preset whale optimization algorithm. The position of each individual in the initial population represents a set of VMD parameters to be optimized. The data filtering module is used to perform a first VMD decomposition on the vibration signal based on the VMD parameters of individuals in the initial population, and to obtain signal envelope correlation features by combining Hilbert transformation to calculate a first fitness, and to filter the initial population to determine a potential solution population based on the first fitness. The data optimization module is used to perform a second VMD decomposition on the vibration signal based on the VMD parameters corresponding to individuals in the potential solution population, calculate the fitness using fault impact related characteristic indicators, and then optimize the potential solution population by combining the population update mechanism of the preset whale optimization algorithm and the global optimal individual perturbation strategy to obtain the global optimal VMD parameters. The fault diagnosis module is used to perform a second VMD decomposition on the vibration signal based on the globally optimal VMD parameters to obtain the intrinsic mode functions, and to filter out sensitive intrinsic multimodal function components through multi-scale permutation entropy, so as to determine fault information based on the sensitive intrinsic multimodal function components and the theoretical characteristics of the robot reducer.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the steps of the industrial robot reducer fault diagnosis method as described in any one of claims 1 to 5.

8. An electronic device, characterized in that, It includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the industrial robot reducer fault diagnosis method as described in any one of claims 1-5.

Images