Industrial robot real-time fault detection method and system

Abstract

The invention discloses a real-time fault detection method and system for an industrial robot, and the method comprises the steps: obtaining state data of the industrial robot in previous normal work, and calculating an average value [mu] and a covariance matrix [sigma] of the state data; the method comprises the following steps: acquiring state data of an operating industrial robot in real time every s seconds, taking W seconds as an acquisition period, and acquiring W / s times in the acquisition period; calculating a density estimation value of each piece of state data in the current acquisition period through multivariate Gaussian distribution obeying an average value mu and a covariance matrix sigma; and finally, according to the density estimation value of each piece of state data in the current acquisition period, judging whether the current industrial robot breaks down or not. According to the method, real-time fault detection of the industrial robot is realized, the computing resource overhead is low, high-accuracy fault detection can be realized without huge deep learning model training, and the probability of occurrence of false alarms can be reduced by using an acquisition period.

Description

Technical Field

[0001] The present invention relates to the field of industrial robot fault detection, and in particular to a real-time fault detection method and system for an industrial robot. Background Art

[0002] Industrial robots are highly integrated complex systems that integrate automation, machinery, and corresponding embedded software and hardware. They usually have multi-joint manipulators or multi-degree-of-freedom machine devices that can replace humans in performing dangerous, complex, or mechanically repetitive tasks. Given their excellent operating accuracy and the characteristics of not requiring rest, industrial robots have been widely used in various industrial fields such as electronics, logistics, and chemicals.

[0003] However, due to the highly integrated and complex nature of industrial robot systems, some uncontrollable and unpredictable failures may occur during the operation of working robots. These failures may cause damage to the equipment at the very least, or threaten the lives of surrounding workers at the worst, and may even lead to catastrophic accidents.

[0004] Therefore, it is very important to effectively detect the operation process of industrial robots in real time in order to eliminate abnormal behaviors in time and establish fault emergency stop devices.

[0005] In the current field of industrial robot fault detection, there are fault detection methods based on computer vision and fault monitoring methods based on deep learning models. Although they have achieved a certain degree of accuracy, they require a large amount of computing resource overhead, which will lead to a long delay in fault detection and cannot achieve true real-time fault detection. On the other hand, huge models require a large number of training samples and are difficult to train.

[0006] The fault monitoring method based on simple threshold judgment, although simple in method and with low computational overhead, has low detection accuracy and cannot perform multi-dimensional fault detection by combining multiple features. Summary of the invention

[0007] In view of the above-mentioned technical deficiencies, the purpose of the present invention is to provide a real-time fault detection method and system for industrial robots, aiming to solve the problems of large resource overhead, large delay, difficult operation, and inability to combine multiple features for multi-dimensional fault detection in the real-time fault detection method of industrial robots.

[0008] In view of the above problems, the present application provides a real-time fault detection method and system for an industrial robot.

[0009] The first aspect disclosed in the present application provides a real-time fault detection method for an industrial robot, the method comprising the following steps: S1: Obtain the state data of the industrial robot under previous normal operation, and calculate the average value μ and covariance matrix Σ of these state data; S2: For the industrial robot in operation, the status data is collected in real time every s seconds, with W seconds as a collection cycle, where W / s data are collected in one collection cycle; S3: Calculate the density estimation value of each state data in the current acquisition period through a multivariate Gaussian distribution that obeys the mean value μ and the covariance matrix Σ; S4: Determine whether the current industrial robot fails based on the density estimation value of each state data in the current acquisition cycle.

[0010] Preferably, the status data consists of temperature values ​​of each motor, torque values ​​of each joint, noise decibel values, vibration acceleration values, voltage values ​​and current values.

[0011] Preferably, the S1 specifically comprises the following steps: S101: Acquire M state data of the industrial robot in previous normal working state, where M≥6000 and includes all normal working scenes of the industrial robot; S102: Calculate the average value μ and covariance matrix Σ of M state data using equations (1) and (2): Formula (1) Formula (2) Among them, x (m) The state vector is composed of the values ​​of the m-th state data arranged in sequence, m=1,2, ...,M.

[0012] Preferably, the S3 specifically includes the following steps: S301: According to the mean value μ and the covariance matrix Σ, a multivariate Gaussian distribution that obeys the mean value μ and the covariance matrix Σ is set N (μ,Σ); S302: Arrange the values ​​of each state data in the current acquisition cycle in order to form a state vector x of each state data (i) , where i=1,2, ... ,W / s; S303: Calculate the multivariate Gaussian distribution using formula (3) N (μ,Σ) on the state vector x (i) The density estimate p(x (i) ;μ,Σ): Formula (3) Among them, |Σ| represents the determinant of the covariance matrix, Σ -1represents the inverse of the covariance matrix.

[0013] Preferably, the S4 specifically includes the following steps: S401: Count the number of density estimation values ​​of state data in the current acquisition cycle that are less than the fault judgment threshold, denoted as g; S402: If g ≥ 0.8·W / s, it is determined that a fault occurs and the industrial robot is controlled to stop urgently. Otherwise, S403 is executed; S403: If g ≥ 0.4·W / s, record suspected fault information once and execute S2; otherwise, execute S404; S404: If g<0.4·W / s, it is determined to be a normal working state and S2 is executed.

[0014] Preferably, the fault determination threshold is the minimum value of the density estimation value of the state data under previous normal operation.

[0015] The second aspect disclosed in the present application provides a real-time fault detection system for an industrial robot, the system being used for the above-mentioned real-time fault detection method for an industrial robot, the system comprising: A parameter estimation module, wherein the parameter estimation module is used to obtain state data of the industrial robot under previous normal operation and calculate the average value μ and covariance matrix Σ of the state data; The acquisition module is used to collect status data of the industrial robot in operation every s seconds in real time, with W seconds as a collection cycle, wherein W / s times are collected in one collection cycle; A density estimation module, wherein the density estimation module calculates a density estimation value of each state data in the current acquisition cycle through a multivariate Gaussian distribution that obeys a mean value μ and a covariance matrix Σ; A fault detection module is used to determine whether a fault occurs in the current industrial robot according to a density estimation value of each state data in a current acquisition cycle.

[0016] According to a third aspect of the present application, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the following steps are implemented: S1: Obtain the state data of the industrial robot under previous normal operation, and calculate the average value μ and covariance matrix Σ of these state data; S2: For the industrial robot in operation, the status data is collected in real time every s seconds, with W seconds as a collection cycle, where W / s data are collected in one collection cycle; S3: Calculate the density estimation value of each state data in the current acquisition period through a multivariate Gaussian distribution that obeys the mean value μ and the covariance matrix Σ; S4: Determine whether the current industrial robot fails based on the density estimation value of each state data in the current acquisition cycle.

[0017] According to a fourth aspect of the present application, a computer-readable storage medium is provided, on which a computer program is stored, and when the computer program is executed by a processor, the following steps are implemented: S1: Obtain the state data of the industrial robot under previous normal operation, and calculate the average value μ and covariance matrix Σ of these state data; S2: For the industrial robot in operation, the status data is collected in real time every s seconds, with W seconds as a collection cycle, where W / s data are collected in one collection cycle; S3: Calculate the density estimation value of each state data in the current acquisition period through a multivariate Gaussian distribution that obeys the mean value μ and the covariance matrix Σ; S4: Determine whether the current industrial robot fails based on the density estimation value of each state data in the current acquisition cycle.

[0018] According to a fifth aspect of the present application, a computer program product is provided, including a computer program or instructions, wherein when the computer program or instructions are executed by a processor, the following steps are implemented: S1: Obtain the state data of the industrial robot under previous normal operation, and calculate the average value μ and covariance matrix Σ of these state data; S2: For the industrial robot in operation, the status data is collected in real time every s seconds, with W seconds as a collection cycle, where W / s data are collected in one collection cycle; S3: Calculate the density estimation value of each state data in the current acquisition period through a multivariate Gaussian distribution that obeys the mean value μ and the covariance matrix Σ; S4: Determine whether the current industrial robot fails based on the density estimation value of each state data in the current acquisition cycle.

[0019] The beneficial effects of the present invention are: (1) It realizes real-time fault detection of industrial robots with low computing resource overhead and low detection latency; (2) There is no need to train large and complex deep learning models, and the system is easy to launch; (3) Using multivariate Gaussian distribution to estimate the density of the probability model can combine multiple features to perform multi-dimensional fault detection, with high accuracy of fault detection; (4) The concept of acquisition cycle is introduced to prevent frequent emergency stops of industrial robots caused by fault detection methods and system false alarms. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0021] Figure 1 A flowchart of a real-time fault detection method for an industrial robot provided in an embodiment of the present application.

[0022] Figure 2 A schematic diagram of the structure of a real-time fault detection system for an industrial robot provided in an embodiment of the present application.

[0023] Description of the reference numerals: parameter estimation module 10 , acquisition module 20 , density estimation module 30 , fault detection module 40 . DETAILED DESCRIPTION

[0024] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0025] like Figure 1 As shown, an embodiment of the present application provides a real-time fault detection method for an industrial robot, the method comprising the following steps: S1: Obtain the state data of the industrial robot under previous normal operation, and calculate the average value μ and covariance matrix Σ of these state data; S2: For the industrial robot in operation, the status data is collected in real time every s seconds, with W seconds as a collection cycle, where W / s data are collected in one collection cycle; S3: Calculate the density estimation value of each state data in the current acquisition period through a multivariate Gaussian distribution that obeys the mean value μ and the covariance matrix Σ; S4: Determine whether the current industrial robot fails based on the density estimation value of each state data in the current acquisition cycle.

[0026] Furthermore, the status data consists of temperature values ​​of each motor, torque values ​​of each joint, noise decibel values, vibration acceleration values, voltage values ​​and current values.

[0027] Furthermore, the S1 specifically includes the following steps: S101: Acquire M state data of the industrial robot in previous normal working state, where M is required to be ≥ 6000 and all normal working scenes of the industrial robot must be included; S102: Calculate the average value μ and covariance matrix Σ of M state data using equations (1) and (2): Formula (1) Formula (2) Among them, x (m) The state vector is composed of the values ​​of the m-th state data arranged in sequence, m=1,2, ...,M.

[0028] Specifically, when the industrial robot is working normally, the state data of the industrial robot is obtained in batches. Some rare special scenes or even extreme scenes can also be artificially created to obtain the state data of the industrial robot still working normally in this scene, so as to increase the robustness of the real-time fault detection method of the industrial robot. The M state data obtained are processed by Python or MATLAB, and the values ​​contained in each state data are arranged in order to generate the state vector corresponding to each state data. The code is written to execute the calculation method of formula (1) and formula (2) to calculate the average value μ and covariance matrix Σ of the M state data. On the other hand, the PyTorch framework or TensorFlow framework can also be used to arrange the values ​​contained in each state data in order to generate a one-dimensional tensor corresponding to each state data, and the average value μ and covariance matrix Σ of the M state data are directly calculated through the interface provided by the framework. The average value μ can also be called mathematical expectation, and the covariance matrix Σ is a symmetric positive definite matrix, which is a natural generalization of the variance.

[0029] Specifically, the present invention requires that the value of M be greater than the number of values ​​contained in the state data, otherwise the covariance matrix Σ will be a singular matrix whose rank is not full, resulting in Σ being irreversible and unable to perform the next step of calculation. On the other hand, the reason for requiring M ≥ 6000 is to eliminate x as much as possible. (m) ,m=1,2, ... ,The proportion of linearly related vectors in M ​​prevents the irreversible situation of Σ and ensures that the next step of calculation can proceed normally.

[0030] Furthermore, the S3 specifically includes the following steps: S301: According to the mean value μ and the covariance matrix Σ, a multivariate Gaussian distribution that obeys the mean value μ and the covariance matrix Σ is set N (μ,Σ); S302: Arrange the values ​​of each state data in the current acquisition cycle in order to form a state vector x of each state data (i) , where i=1,2, ... ,W / s; S303: Calculate the multivariate Gaussian distribution using formula (3) N (μ,Σ) on the state vector x (i) The density estimate p(x (i) ;μ,Σ): Formula (3) Among them, |Σ| represents the determinant of the covariance matrix, Σ -1 represents the inverse of the covariance matrix.

[0031] Specifically, the determinant of the covariance matrix |Σ| can be solved by writing code through recursion or cofactor expansion, or directly through the det() function provided by Python, MATLAB or PyTorch framework. To find the inverse matrix Σ of the covariance matrix Σ -1 You can write code to solve it using Gaussian elimination, adjoint matrix method, or LU decomposition. You can also solve it directly using the inv() function provided by Python, MATLAB, or TensorFlow framework, or the inverse() function provided by PyTorch framework.

[0032] Specifically, multivariate Gaussian distribution, also known as multivariate normal distribution, is a common continuous probability distribution in statistics, used to describe the joint distribution of multiple random variables. (1) ,x (2) , ... ,x (M) are random variables, and they obey a multivariate Gaussian distribution. The probability density function of the distribution is shown in formula (3). The key to the present invention is to establish a p(x (i) ;μ,Σ) probability model, formula (3) actually corresponds to the independent assumption of each value of a state data, but in actual operation, regardless of whether these values ​​are independent, the effect of the method provided by the present invention is relatively good. Estimation p(x (i) ;μ,Σ) is called the density estimation problem. The density estimation value is used to determine whether the real-time collected status data is abnormal data. If the density estimation value is very small, then this item is marked as a fault.

[0033] Furthermore, the S4 specifically includes the following steps: S401: Count the number of density estimation values ​​of state data in the current acquisition cycle that are less than the fault judgment threshold, denoted as g; S402: If g ≥ 0.8·W / s, it is determined that a fault occurs and the industrial robot is controlled to stop urgently. Otherwise, S403 is executed; S403: If g ≥ 0.4·W / s, record suspected fault information once and execute S2; otherwise, execute S404; S404: If g<0.4·W / s, it is determined to be a normal working state and S2 is executed.

[0034] Specifically, it is counted how many density estimation values ​​of the status data in the current collection cycle are less than the fault judgment threshold. If more than 80% of the density estimation values ​​of the status data in the current collection cycle are less than the fault judgment threshold, it is directly determined that a fault has occurred, and the industrial robot should be controlled to stop urgently; if more than 40% of the density estimation values ​​of the status data in the current collection cycle are less than the fault judgment threshold, but less than 80%, it is determined that a suspected fault has occurred, and the status data in the current collection cycle is recorded in the database as a warning message, which should also include the current timestamp; if less than 40% of the density estimation values ​​of the status data in the current collection cycle are less than the fault judgment threshold, it is determined that the current working state is normal and no fault has occurred.

[0035] Specifically, by introducing the concept of acquisition cycle, whether a fault occurs is determined by the number of density estimation values ​​of state data in the current acquisition cycle that are less than the fault judgment threshold. This can reduce misjudgments caused by the rarity of certain special scenarios, avoid frequent emergency stops of industrial robots, improve the accuracy of real-time fault detection, and reduce the misjudgment rate.

[0036] Furthermore, the fault determination threshold is the minimum value of the density estimation value of the state data under previous normal operation.

[0037] Specifically, assuming that the fault judgment threshold is t, the calculation method of t is shown in formula (4): Formula (4) Among them, min represents the minimum function. In actual operation, there are other ways to calculate the fault judgment threshold, such as the cross-validation set method, the method of finding the maximum value of the abnormal sample density estimation, etc., but these methods will increase the complexity of the calculation and are not helpful to improve the accuracy of real-time fault detection of industrial robots.

[0038] In summary, the real-time fault detection method for an industrial robot provided by the embodiment of the present application has the following technical effects: (1) It realizes real-time fault detection of industrial robots with low computing resource overhead and low detection latency; (2) There is no need to train large and complex deep learning models, and the system is easy to launch; (3) Using multivariate Gaussian distribution to estimate the density of the probability model can combine multiple features to perform multi-dimensional fault detection, with high accuracy of fault detection; (4) The concept of acquisition cycle is introduced to prevent frequent emergency stops of industrial robots caused by fault detection methods and system false alarms.

[0039] Based on the same inventive concept as the real-time fault detection method for an industrial robot in the aforementioned embodiment, Figure 2 As shown, the present application provides a real-time fault detection system for an industrial robot, the system comprising: A parameter estimation module 10, wherein the parameter estimation module 10 is used to obtain state data of the industrial robot under previous normal operation, and calculate the average value μ and covariance matrix Σ of the state data; The acquisition module 20 is used to collect status data of the industrial robot in operation in real time every s seconds, with W seconds as a collection cycle, wherein W / s data are collected in one collection cycle.

[0040] A density estimation module 30, wherein the density estimation module 30 calculates a density estimation value of each state data in the current acquisition cycle through a multivariate Gaussian distribution that obeys a mean value μ and a covariance matrix Σ; A fault detection module 40, wherein the fault detection module 40 is used to determine whether a fault occurs in the current industrial robot according to a density estimation value of each state data in a current acquisition cycle; Furthermore, the parameter estimation module 10 includes the following submodules: Data acquisition submodule: The data acquisition submodule is used to obtain M state data of the industrial robot under previous normal operation, wherein M is required to be ≥ 6000 and all normal operation scenes of the industrial robot must be included; Parameter calculation submodule: The parameter calculation submodule is used to calculate the average value μ and covariance matrix Σ of M state data using formula (1) and formula (2): Formula (1) Formula (2) Among them, x (m) The state vector is composed of the values ​​of the m-th state data arranged in sequence, m=1,2, ...,M.

[0041] Furthermore, the density estimation module 30 includes the following submodules: Distribution setting submodule: The distribution setting submodule is used to set a multivariate Gaussian distribution that obeys the mean value μ and the covariance matrix Σ according to the mean value μ and the covariance matrix Σ N (μ,Σ); Vector formation submodule: The vector formation submodule is used to arrange the values ​​of each state data in the current acquisition cycle in order to form a state vector x of each state data (i) , where i=1,2, ... ,W / s; Numerical calculation submodule: The numerical calculation submodule is used to calculate the multivariate Gaussian distribution using formula (3). N (μ,Σ) on the state vector x (i) The density estimate p(x (i) ;μ,Σ): Formula (3) Among them, |Σ| represents the determinant of the covariance matrix, Σ -1 represents the inverse of the covariance matrix.

[0042] Furthermore, the fault detection module 40 includes the following submodules: Statistics submodule: The statistics submodule is used to count the number of density estimation values ​​of state data in the current acquisition cycle that are less than the fault judgment threshold, denoted as g; Determination submodule: The determination submodule is used to determine whether the working state of the industrial robot is normal according to the size relationship between g and W / s.

[0043] Through the above detailed description of a real-time fault detection method for an industrial robot in this specification, those skilled in the art can clearly understand a real-time fault detection system for an industrial robot in this embodiment. Since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part description.

[0044] In one embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and when the processor executes the computer program, the following steps are implemented: S1: Obtain the state data of the industrial robot under previous normal operation, and calculate the average value μ and covariance matrix Σ of these state data; S2: For the industrial robot in operation, the status data is collected in real time every s seconds, with W seconds as a collection cycle, where W / s data are collected in one collection cycle; S3: Calculate the density estimation value of each state data in the current acquisition period through a multivariate Gaussian distribution that obeys the mean value μ and the covariance matrix Σ; S4: Determine whether the current industrial robot fails based on the density estimation value of each state data in the current acquisition cycle.

[0045] In one embodiment, a computer readable storage medium is provided, on which a computer program is stored, and when the computer program is executed by a processor, the following steps are implemented: S1: Obtain the state data of the industrial robot under previous normal operation, and calculate the average value μ and covariance matrix Σ of these state data; S2: For the industrial robot in operation, the status data is collected in real time every s seconds, with W seconds as a collection cycle, where W / s data are collected in one collection cycle; S3: Calculate the density estimation value of each state data in the current acquisition period through a multivariate Gaussian distribution that obeys the mean value μ and the covariance matrix Σ; S4: Determine whether the current industrial robot fails based on the density estimation value of each state data in the current acquisition cycle.

[0046] In one embodiment, a computer program product is provided, including a computer program or instructions, which, when executed by a processor, implements the following steps: S1: Obtain the state data of the industrial robot under previous normal operation, and calculate the average value μ and covariance matrix Σ of these state data; S2: For the industrial robot in operation, the status data is collected in real time every s seconds, with W seconds as a collection cycle, where W / s data are collected in one collection cycle; S3: Calculate the density estimation value of each state data in the current acquisition period through a multivariate Gaussian distribution that obeys the mean value μ and the covariance matrix Σ; S4: Determine whether the current industrial robot fails based on the density estimation value of each state data in the current acquisition cycle.

[0047] The technical features of the above embodiments may be arbitrarily combined. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0048] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A real-time fault detection method for an industrial robot, characterized in that: The method comprises the following steps: S1: Obtain the state data of the industrial robot under previous normal operation, and calculate the average value μ and covariance matrix Σ of these state data; S2: For the industrial robot in operation, the status data is collected in real time every s seconds, with W seconds as a collection cycle, where W / s data are collected in one collection cycle; S3: Calculate the density estimation value of each state data in the current acquisition period through a multivariate Gaussian distribution that obeys the mean value μ and the covariance matrix Σ; S4: Determine whether the current industrial robot fails based on the density estimation value of each state data in the current acquisition cycle.

2. A real-time fault detection method for an industrial robot as claimed in claim 1, characterized in that: The state data consists of the temperature value of each motor, the torque value of each joint, the noise decibel value, the vibration acceleration value, the voltage value and the current value.

3. The real-time fault detection method for an industrial robot according to claim 1, characterized in that: The S1 specifically includes the following steps: S101: Acquire M state data of the industrial robot in previous normal working state, where M≥6000 and includes all normal working scenes of the industrial robot; S102: Calculate the average value μ and covariance matrix Σ of M state data using equations (1) and (2): Formula (1) Formula (2) Among them, x (m) The state vector is composed of the values ​​of the m-th state data arranged in sequence, m=1,2, ...,M.

4. The real-time fault detection method for an industrial robot according to claim 1, characterized in that: The S3 specifically includes the following steps: S301: According to the mean value μ and the covariance matrix Σ, a multivariate Gaussian distribution that obeys the mean value μ and the covariance matrix Σ is set N (μ,Σ); S302: Arrange the values ​​of each state data in the current acquisition cycle in order to form a state vector x of each state data (i) , where i=1,2, ... ,W / s; S303: Calculate the multivariate Gaussian distribution using formula (3) N (μ,Σ) on the state vector x (i) The density estimate p(x (i) ;μ,Σ): Formula (3) Among them, |Σ| represents the determinant of the covariance matrix, Σ -1 represents the inverse of the covariance matrix.

5. The real-time fault detection method for an industrial robot according to claim 1, characterized in that: The S4 specifically comprises the following steps: S401: Count the number of density estimation values ​​of state data in the current acquisition cycle that are less than the fault judgment threshold, denoted as g; S402: If g ≥ 0.8·W / s, it is determined that a fault occurs and the industrial robot is controlled to stop urgently. Otherwise, S403 is executed; S403: If g ≥ 0.4·W / s, record suspected fault information once and execute S2; otherwise, execute S404; S404: If g<0.4·W / s, it is determined to be a normal working state and S2 is executed.

6. A real-time fault detection method for an industrial robot as claimed in claim 5, characterized in that: The fault determination threshold is the minimum value of the density estimation value of the state data under previous normal operation.

7. A real-time fault detection system for industrial robots, characterized in that: The system comprises: A parameter estimation module, wherein the parameter estimation module is used to obtain state data of the industrial robot under previous normal operation and calculate the average value μ and covariance matrix Σ of the state data; The acquisition module is used to collect status data of the industrial robot in operation every s seconds in real time, with W seconds as a collection cycle, wherein W / s times are collected in one collection cycle; A density estimation module, wherein the density estimation module calculates a density estimation value of each state data in the current acquisition cycle through a multivariate Gaussian distribution that obeys a mean value μ and a covariance matrix Σ; A fault detection module is used to determine whether a fault occurs in the current industrial robot according to a density estimation value of each state data in a current acquisition cycle.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the real-time fault detection method for an industrial robot according to any one of claims 1 to 6 are implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of a real-time fault detection method for an industrial robot according to any one of claims 1 to 6 are implemented.

10. A computer program product comprising a computer program or instructions, characterized in that When the computer program or instruction is executed by a processor, the steps of a real-time fault detection method for an industrial robot according to any one of claims 1 to 6 are implemented.

Images