Mechanical arm fault diagnosis method and device, computer device, and storage medium

Abstract

The application relates to a mechanical arm fault diagnosis method and device, computer equipment and a storage medium. The method comprises the following steps: based on normal training torque data and corresponding multiple abnormal training torque data in a preset database, a training torque matrix is constructed; according to the training torque matrix, an initial fault diagnosis model is trained to determine a target fault diagnosis model; actual difference data groups of the mechanical arm are acquired, and an actual torque matrix is determined; based on the target fault diagnosis model, fault prediction is performed according to the actual torque matrix to determine the fault state of the mechanical arm. The method can comprehensively diagnose the fault of the mechanical arm.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of mechanical arms, in particular to a mechanical arm fault diagnosis method and device, computer equipment and a storage medium. BACKGROUND

[0002] In the process of disassembling equipment in special scenarios such as the nuclear industry, in order to protect the health of personnel, a mechanical arm is generally controlled by a remote control to disassemble the equipment. The PLC directly collects mechanical arm operation data such as torque, speed and displacement signals for the evaluation of the running state of the mechanical arm and the determination of the health of the equipment.

[0003] With the development of mechanical arm technology, mechanical arm fault diagnosis technology has emerged, which models fault diagnosis by directly collecting mechanical arm data to monitor the running state of the mechanical arm. However, this method lacks enough reliable fault samples, and the data extracted is usually single-dimensional data, making it impossible to comprehensively diagnose faults in the mechanical arm.

[0004] There is an urgent need for a solution that can comprehensively diagnose faults in a mechanical arm. SUMMARY

[0005] Therefore, it is necessary to provide a mechanical arm fault diagnosis method, device, computer equipment and storage medium that can comprehensively diagnose faults in a mechanical arm to solve the above technical problems.

[0006] In a first aspect, the present application provides a mechanical arm fault diagnosis method. The method comprises: constructing a training torque matrix based on normal training torque data and corresponding multiple abnormal training torque data in a preset database; training a preset initial fault diagnosis model according to the training torque matrix to determine a target fault diagnosis model; obtaining an actual difference data set of the mechanical arm and determining an actual torque matrix; and performing fault prediction based on the target fault diagnosis model according to the actual torque matrix to determine the fault state of the mechanical arm.

[0007] In one embodiment, the multiple abnormal training torque data includes abnormal training torque data corresponding to multiple fault types; the timestamps of the normal training torque data correspond to the timestamps of each abnormal training torque data; the training torque matrix is constructed based on the normal training torque data and the corresponding multiple abnormal training torque data in the preset database by: subtracting each abnormal training torque data from the normal training torque data to obtain a torque residual sequence corresponding to each abnormal training torque data; determining the training torque matrix according to the torque residual sequence corresponding to each abnormal training torque data; and each torque residual sequence in the training torque matrix corresponds to a fault type label.

[0008] In one of the embodiments, the determining the training torque matrix according to the torque residual sequence corresponding to each of the abnormal training torque data comprises: performing linear fitting on the target torque residual sequence corresponding to the target abnormal training torque data to determine a plurality of torque fitting line segments; the target abnormal training torque data is any one of the abnormal training torque data; and determining the training torque matrix according to the preset trend coding rule and the plurality of torque fitting line segments of each of the abnormal training torque data.

[0009] In one of the embodiments, the determining the training torque matrix according to the preset trend coding rule and the plurality of torque fitting line segments of each of the abnormal training torque data comprises: determining a first trend parameter of the target torque fitting line segment according to a last residual value of the target torque fitting line segment and a last residual value of a previous torque fitting line segment of the target torque fitting line segment; determining a second trend parameter of the target torque fitting line segment according to a first residual value of the target torque fitting line segment and the last residual value of the previous torque fitting line segment of the target torque fitting line segment; determining a third trend parameter of the target torque fitting line segment according to the first residual value of the target torque fitting line segment and a last residual value of the target torque fitting line segment; and determining the training torque matrix according to the first trend parameter, the second trend parameter and the third trend parameter based on the preset first threshold and the second threshold.

[0010] In one of the embodiments, the determining the training torque matrix according to the first trend parameter, the second trend parameter and the third trend parameter based on the preset first threshold and the second threshold comprises: determining a first number or a second number of the target torque fitting line segment according to the first trend parameter, the second trend parameter and the third trend parameter based on the preset first threshold and the second threshold; determining a third number or a fourth number of the target torque fitting line segment according to the second trend parameter and the third trend parameter based on the preset first threshold and the second threshold; determining a fifth number or a sixth number of the target torque fitting line segment according to the second trend parameter and the third trend parameter based on the preset first threshold and the second threshold; and determining the training torque matrix according to the number corresponding to each of the torque fitting line segments.

[0011] In one of the embodiments, the method further comprises: obtaining a group of angular displacements; the angular displacements in the group of angular displacements are end segment angular displacements of the robot arm; based on a preset angular velocity calculation formula and an angular acceleration calculation formula, determining a group of angular velocities and a group of angular accelerations according to the group of angular displacements, a plurality of angular velocity values in the group of angular velocities and a plurality of angular acceleration values in the group of angular accelerations correspond to a plurality of end segment angular displacements in the group of angular displacements; based on a preset torque calculation formula, determining normal training torque data according to the group of angular displacements, the group of angular velocities and the group of angular accelerations; based on a preset first abnormal formula, determining first abnormal training torque data according to the group of angular displacements, the group of angular velocities and the group of angular accelerations; based on a preset second abnormal formula, determining second abnormal training torque data according to the group of angular displacements, the group of angular velocities and the group of angular accelerations; based on a preset third abnormal formula, determining third abnormal training torque data according to the group of angular displacements, the group of angular velocities and the group of angular accelerations; and determining abnormal training torque data according to the first abnormal training torque data, the second abnormal training torque data and the third abnormal training torque data.

[0012] In one of the embodiments, the obtaining the actual difference value data group of the robot arm and determining the actual torque matrix comprises: obtaining a plurality of groups of instructions, the instructions comprising a plurality of angular displacement instructions; the angular displacement instructions are used to control the corresponding joint movement of the robot arm; controlling the robot arm according to the plurality of groups of instructions respectively, and obtaining the actual torque of the robot arm corresponding to each of the groups of instructions to determine an actual torque group; based on a preset angular velocity calculation formula, an angular acceleration calculation formula and a torque calculation formula, determining an ideal torque group according to the plurality of groups of instructions; a plurality of ideal torques in the ideal torque group correspond to a plurality of actual torques in the actual torque group one by one; determining the actual difference value data group by subtracting the ideal torque group from the actual torque group; performing linear fitting according to the actual difference value data group to determine a plurality of fitting line segments; and determining the actual torque matrix according to a preset trend coding rule and the plurality of fitting line segments.

[0013] In a second aspect, the present application further provides a robot arm fault diagnosis device. The device comprises:

[0014] a construction module configured to construct a training torque matrix based on the normal training torque data and the plurality of abnormal training torque data in the preset database; a training module configured to train a preset initial fault diagnosis model according to the training torque matrix to determine a target fault diagnosis model; an obtaining module configured to obtain an actual difference value data group of the robot arm and determine an actual torque matrix; and a determination module configured to perform fault prediction according to the actual torque matrix based on the target fault diagnosis model to determine the fault state of the robot arm.

[0015] In a third aspect, the present application also provides a computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement any method in the first aspect.

[0016] In a fourth aspect, the present application also provides a computer readable storage medium, which stores a computer program, and the computer program is executed by a processor to implement any method in the first aspect.

[0017] The mechanical arm fault diagnosis method, device, computer device and storage medium can comprehensively diagnose the fault of the mechanical arm by constructing a training torque matrix based on normal training torque data and corresponding multiple abnormal training torque data in a preset database, training a preset initial fault diagnosis model according to the training torque matrix to determine a target fault diagnosis model, obtaining actual difference data sets of the mechanical arm to determine an actual torque matrix, and performing fault prediction based on the target fault diagnosis model and the actual torque matrix to determine the fault state of the mechanical arm. BRIEF DESCRIPTION OF DRAWINGS

[0018] Figure 1 An application environment diagram of the mechanical arm fault diagnosis method in an embodiment;

[0019] Figure 2 A flowchart of the mechanical arm fault diagnosis method in an embodiment;

[0020] Figure 3 An architecture diagram of the target fault diagnosis model in an embodiment;

[0021] Figure 4 A logic diagram for determining the training torque matrix in another embodiment;

[0022] Figure 5 A diagram for determining different trends of torque fitting line segments in another embodiment;

[0023] Figure 6 A diagram of the mechanical arm simulation model in another embodiment;

[0024] Figure 7 A system architecture diagram of the mechanical arm fault diagnosis method in another embodiment;

[0025] Figure 8 A flow architecture diagram of the mechanical arm fault diagnosis method in another embodiment;

[0026] Figure 9 A structural block diagram of the mechanical arm fault diagnosis device in an embodiment;

[0027] Figure 10 Figure 1 is a schematic diagram of the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0028] In order to make the purpose, technical solutions and advantages of the present application clearer, the present application will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and not to limit the present application.

[0029] Since the operating environment of the mechanical arm in the nuclear industry site is radioactive, the device can only be controlled in the form of remote operation, and it is difficult to directly detect the abnormal state of the device on site. Especially when the mechanical arm performs a production task, there are often highly radioactive substances in the working environment of the mechanical arm, so the accuracy of the fault reasoning result of the model to the mechanical arm is required to be high. The conventional scheme cannot quickly and effectively extract the features associated with the health status of the mechanical arm because it lacks sufficient reliable fault samples and the extracted analysis data and features are usually single-dimensional and there is a lot of redundant information. Moreover, this method needs to carry out various fault simulation tests on the mechanical arm itself to collect device operation data under various conditions, needs to customize special fault parts, and needs to set up a special test site, consuming a lot of manpower and resources.

[0030] In addition, the conventional fault diagnosis model lacks an effective online discrimination mechanism, cannot timely discover the fault of the mechanical arm, and the abnormality of the fault diagnosis model itself will cause the reliability of the reasoning result to decrease, thereby causing the accuracy of the prediction result to decrease.

[0031] The mechanical arm fault diagnosis method provided by the embodiments of the present application can be applied to the application environment as shown in Figure 1 . The terminal 102 communicates with the server 104 through the network. The data storage system can store the data required to be processed by the server 104. The data storage system can be integrated on the server 104, or can be placed on the cloud or other network servers. The server 104 is used to execute the mechanical arm fault diagnosis method. The terminal 102 can be but is not limited to various different types of mechanical arms. The server 104 can be implemented by an independent server or a server cluster composed of multiple servers.

[0032] In order to solve the above problems, in one embodiment of the present application, as shown in Figure 2 , a mechanical arm fault diagnosis method is provided, comprising the following steps:

[0033] Step 201, based on the normal training torque data and the corresponding multiple abnormal training torque data in the preset database, a training torque matrix is constructed.

[0034] The normal training torque data is torque data of the robot arm when performing an operation in a normal state, and the abnormal training torque data is torque data of the robot arm when performing the operation in an abnormal state. The training torque matrix is a matrix for training the initial fault diagnosis model, which is constructed based on the normal training torque data and a plurality of corresponding abnormal training torque data.

[0035] That is, the normal training torque data and the plurality of corresponding abnormal training torque data are subtracted to obtain a plurality of torque residual sequences, so as to construct the training torque matrix.

[0036] It can be understood that the normal training torque data and the corresponding abnormal training torque data are torque data under the premise of performing the same operation. The torque data of the robot arm when performing the same operation in different abnormal states is also different, so the number of the abnormal training torque data corresponding to the normal training torque data is multiple.

[0037] In other embodiments of the present application, when the abnormal state of the robot arm includes a stuck fault, a free swing, and component aging, the number of the abnormal training torque data corresponding to the normal training torque data is three.

[0038] In addition, although only one normal training torque data and a plurality of corresponding abnormal training torque data are processed in the description of the present embodiment, in actual environment, a plurality of normal training torque data and corresponding abnormal training torque data are processed.

[0039] Step 202, training the preset initial fault diagnosis model according to the training torque matrix to determine a target fault diagnosis model.

[0040] The initial fault diagnosis model is an initial model capable of preliminary fault diagnosis, and the target fault diagnosis model is a fault diagnosis model after training.

[0041] It should be noted that, in an embodiment of the present application, as shown in Figure 3 The target fault diagnosis model is composed of an input layer, a first convolutional layer, a first pooling layer, a second convolutional layer, a second pooling layer, a first full connection layer, a second full connection layer, and an output layer. The convolutional layer activation function selects a Relu activation function, the pooling layer selects a maximum pooling, the full connection layer selects a Relu activation function, and the output layer is temporarily divided into three neurons according to the identification of the fault type, which are [0, 0, 0] to represent that the robot arm is normal, [1, 0, 0] to represent that the robot arm joint is stuck, [0, 1, 0] to represent that the robot arm joint appears free swing, and [0, 0, 1] to represent that the robot arm appears electrical component aging. In other embodiments of the present application, identification of other fault types can be achieved by increasing the output layer and inputting new abnormal training torque data to retrain the deep learning model.

[0042] Step 203, obtaining an actual difference data set of the mechanical arm, and determining an actual torque matrix.

[0043] The actual difference data set is a data set of the mechanical arm in actual use. The actual torque matrix is a matrix for representing the actual state of the mechanical arm.

[0044] Step 204, based on the target fault diagnosis model, performing fault prediction according to the actual torque matrix, and determining the fault state of the mechanical arm.

[0045] That is, inputting the actual torque matrix into the target fault diagnosis model to perform fault prediction, thereby determining the fault state of the mechanical arm.

[0046] In the above mechanical arm fault diagnosis method, the training torque matrix is constructed based on the normal training torque data and the corresponding plurality of abnormal training torque data in the preset database; the initial fault diagnosis model is trained according to the training torque matrix to determine the target fault diagnosis model; then the actual difference data set of the mechanical arm is obtained to determine the actual torque matrix; finally, based on the target fault diagnosis model, the fault prediction is performed according to the actual torque matrix to determine the fault state of the mechanical arm, so that the fault diagnosis of the mechanical arm can be comprehensively performed.

[0047] In other embodiments of the present application, the plurality of abnormal training torque data includes abnormal training torque data corresponding to a plurality of fault types; the timestamps of the normal training torque data correspond to the timestamps of each abnormal training torque data; and the training torque matrix is constructed based on the normal training torque data and the corresponding plurality of abnormal training torque data in the preset database, including:

[0048] Step 1, respectively subtracting the normal training torque data from the abnormal training torque data of each fault type to obtain a torque residual sequence corresponding to each abnormal training torque data.

[0049] Exemplarily, the plurality of fault types includes a stuck fault, a free swing, and a component aging.

[0050] It should be noted that the timestamps of the normal training torque data correspond to the timestamps of each abnormal training torque data, that is, the sampling time of the normal training torque data and the sampling time of the corresponding plurality of abnormal training torque data are the same time.

[0051] It should be noted that the number of normal training torque data is multiple, the number of abnormal training torque data is also multiple, and the number of each normal training torque data is the same as the number of abnormal training torque data of each fault type. The normal training torque data is respectively subtracted from the abnormal training torque data of each fault type to obtain the torque residual sequence corresponding to each abnormal training torque data. Specifically, each normal training torque data is subtracted from the abnormal training torque data corresponding to the timestamp of each fault type, thereby obtaining the residual value between each normal training torque data and the abnormal training torque data of each fault type, thereby serving as the torque residual sequence corresponding to each abnormal training torque data.

[0052] Among the torque residual sequence corresponding to a type of fault type, all residual values between each normal training torque data and the abnormal training torque data of this type of fault type are included.

[0053] Step 2, determine the training torque matrix according to the torque residual sequence corresponding to each abnormal training torque data.

[0054] That is, the torque residual sequence is linearly fitted, and the training torque matrix is determined according to the preset trend coding rule. Each torque residual sequence in the training torque matrix corresponds to a fault type label. It should be noted that each row of the training torque matrix corresponds to the residual value of a fault type at different times, and each column corresponds to the residual value corresponding to different fault types at the same time.

[0055] In other embodiments of the present application, determining the training torque matrix according to the torque residual sequence corresponding to each abnormal training torque data comprises:

[0056] Step 1, linearly fitting the target torque residual sequence corresponding to the target abnormal training torque data to determine a plurality of torque fitting line segments.

[0057] That is, the target torque residual sequence corresponding to the target abnormal training torque data is linearly fitted, so that the plurality of residual values in the target torque residual sequence and the corresponding torque fitting line segment are within a predetermined range, and the plurality of torque fitting line segments are not connected.

[0058] It should be noted that in the process of linear fitting, the first three residual values in the target torque residual sequence are first fitted, and it is judged whether the gap between the obtained torque fitting line segment and the first three residual values in the target torque residual sequence is within the preset range. If yes, the first four residual values in the target torque residual sequence are re-fitted, and it is re-judged whether the gap between the obtained torque fitting line segment and the first four residual values in the target torque residual sequence is within the preset range. If no, the torque fitting line segment fitted by the first two residual values is saved, and the third and fourth residual values in the target torque residual sequence are repeated for the above steps until multiple torque fitting line segments that are not connected to each other are obtained.

[0059] For example, when fitting the first n residual values, the first n residual values in the target torque residual sequence are first fitted, and it is judged whether the gap between the obtained torque fitting line segment and the first n residual values in the target torque residual sequence is within the preset range. If yes, the first n+1 residual values in the target torque residual sequence are re-fitted, and it is re-judged whether the gap between the obtained torque fitting line segment and the first n+1 residual values in the target torque residual sequence is within the preset range. If no, the torque fitting line segment fitted by the first n-1 residual values is saved, and the n-th and n+1-th residual values in the target torque residual sequence are fitted, and it is judged whether the gap between the obtained torque fitting line segment and the n-th and n+1-th residual values in the target torque residual sequence is within the preset range, until multiple torque fitting line segments that are not connected to each other are obtained.

[0060] The target abnormal training torque data is any one of the abnormal training torque data.

[0061] It should be noted that in the present embodiment, although only the target abnormal training torque data corresponding to the target torque residual sequence is executed, in actual application, the torque residual sequence corresponding to all abnormal training torque data is executed.

[0062] Step 2, determining the training torque matrix according to the preset trend coding rule and the multiple torque fitting line segments of each abnormal training torque data.

[0063] That is, the first trend parameter, the second trend parameter and the third trend parameter of each torque fitting line segment are calculated, and the training torque matrix is finally determined based on the preset trend coding rule.

[0064] In other embodiments of the present application, the determination of the training torque matrix according to the preset trend coding rule and the multiple torque fitting line segments of each abnormal training torque data comprises:

[0065] Step 1, determining a first trend parameter of the target torque fitting line segment according to a last residual value of the target torque fitting line segment and a last residual value of a previous torque fitting line segment of the target torque fitting line segment.

[0066] That is, calculating a difference value between the last residual value of the target torque fitting line segment and the last residual value of the previous torque fitting line segment of the target torque fitting line segment as the first trend parameter of the target torque fitting line segment.

[0067] Step 2, determining a second trend parameter of the target torque fitting line segment according to a first residual value of the target torque fitting line segment and the last residual value of the previous torque fitting line segment of the target torque fitting line segment.

[0068] That is, calculating a difference value between the first residual value of the target torque fitting line segment and the last residual value of the previous torque fitting line segment of the target torque fitting line segment as the second trend parameter of the target torque fitting line segment.

[0069] Step 3, determining a third trend parameter of the target torque fitting line segment according to the first residual value of the target torque fitting line segment and the last residual value of the target torque fitting line segment.

[0070] That is, calculating a difference value between the first residual value of the target torque fitting line segment and the last residual value of the target torque fitting line segment as the third trend parameter of the target torque fitting line segment.

[0071] Step 4, determining the training torque matrix according to the first trend parameter, the second trend parameter and the third trend parameter based on the preset first threshold value and the second threshold value.

[0072] That is, encoding according to the first trend parameter, the second trend parameter and the third trend parameter of each torque fitting line segment based on the preset trend encoding rule, the preset first threshold value and the second threshold value, so as to determine the training torque matrix.

[0073] In other embodiments of the present application, the determination of the training torque matrix according to the first trend parameter, the second trend parameter and the third trend parameter based on the preset first threshold value and the second threshold value comprises:

[0074] Step 1, determining that the target torque fitting line segment is numbered as a first number or a second number according to the first trend parameter, the second trend parameter and the third trend parameter based on the preset first threshold value and the second threshold value.

[0075] Step 2, determining that the target torque fitting line segment is numbered as a third number or a fourth number according to the second trend parameter and the third trend parameter based on the preset first threshold value and the second threshold value.

[0076] Step 3, determining the number of the target torque fitting line segment as the fifth number or the sixth number according to the second trend parameter and the third trend parameter based on the preset first threshold value and the second threshold value;

[0077] Step 4, determining the training torque matrix according to the number corresponding to each torque fitting line segment.

[0078] In other embodiments of the present application, as shown in Figure 4 determining the training torque matrix based on the preset first threshold value and the second threshold value according to the first trend parameter, the second trend parameter and the third trend parameter includes:

[0079] Step 1, if the absolute value of the second trend parameter is less than the first threshold value, and the absolute value of the first trend parameter is greater than the second threshold value, and the first trend parameter is greater than zero, then the number of the target torque fitting line segment is the first number.

[0080] That is, if the absolute value of the second trend parameter c d is less than the first threshold value V1, and the absolute value of the first trend parameter c is greater than the second threshold value V2, and the first trend parameter c is greater than zero, then the number of the target torque fitting line segment is the first number 1, which represents that the target torque fitting line segment shows an upward trend compared with the previous torque fitting line segment.

[0081] Step 2, if the absolute value of the second trend parameter is less than the first threshold value, and the absolute value of the first trend parameter is greater than or equal to the second threshold value, and the first trend parameter is less than zero, then the number of the target torque fitting line segment is the second number.

[0082] That is, if the absolute value of the second trend parameter c d is less than the first threshold value V1, and the absolute value of the first trend parameter c is greater than or equal to the second threshold value V2, and the first trend parameter c is less than zero, then the number of the target torque fitting line segment is the second number -1, which represents that the target torque fitting line segment shows a downward trend compared with the previous torque fitting line segment.

[0083] Step 3, if the absolute value of the second trend parameter is greater than or equal to the first threshold value, and the absolute value of the third trend parameter is less than or equal to the second threshold value, and the first trend parameter is greater than zero, then the number of the target torque fitting line segment is the third number.

[0084] That is, if the absolute value of the second trend parameter c d is greater than or equal to the first threshold value V1, and the absolute value of the third trend parameter c s is less than or equal to the second threshold value V2, and the first trend parameter c is greater than zero, then the number of the target torque fitting line segment is the third number 2, which represents that the target torque fitting line segment shows an upward trend compared with the previous torque fitting line segment.

[0085] Step 4, if the absolute value of the second trend parameter is greater than or equal to the first threshold value, and the absolute value of the third trend parameter is less than or equal to the second threshold value, and the first trend parameter is less than zero, then the number of the target torque fitting line segment is a fourth number.

[0086] That is, if the absolute value of the second trend parameter c d is greater than or equal to the first threshold value V1, and the absolute value of the third trend parameter c s is less than or equal to the second threshold value V2, and the first trend parameter c is less than zero, then the number of the target torque fitting line segment is a fourth number -2, which represents that the target torque fitting line segment has a downward trend compared with the previous torque fitting line segment.

[0087] Step 5, if the absolute value of the second trend parameter is greater than or equal to the first threshold value, and the absolute value of the third trend parameter is greater than the second threshold value, and the signs of the second trend parameter and the third trend parameter are the same, and the second trend parameter is greater than zero, then the number of the target torque fitting line segment is the first number.

[0088] That is, if the absolute value of the second trend parameter c d is greater than or equal to the first threshold value V1, and the absolute value of the third trend parameter c s is greater than the second threshold value V2, and the signs of the second trend parameter c d and the third trend parameter c s are the same, and the second trend parameter c d is greater than zero, then the number of the target torque fitting line segment is the first number 1, which represents that the target torque fitting line segment has an upward trend compared with the previous torque fitting line segment.

[0089] Step 6, if the absolute value of the second trend parameter is greater than or equal to the first threshold value, and the absolute value of the third trend parameter is greater than the second threshold value, and the signs of the second trend parameter and the third trend parameter are the same, and the second trend parameter is less than zero, then the number of the target torque fitting line segment is the second number.

[0090] That is, if the absolute value of the second trend parameter c d is greater than or equal to the first threshold value V1, and the absolute value of the third trend parameter c s is greater than the second threshold value V2, and the signs of the second trend parameter c d and the third trend parameter c s are the same, and the second trend parameter c d is less than zero, then the number of the target torque fitting line segment is the second number -1, which represents that the target torque fitting line segment has a downward trend compared with the previous torque fitting line segment.

[0091] Step 7, if the absolute value of the second trend parameter is greater than or equal to the first threshold value, and the absolute value of the third trend parameter is greater than the second threshold value, and the signs of the second trend parameter and the third trend parameter are opposite, and the second trend parameter is greater than zero, then the number of the target torque fitting line segment is a fifth number.

[0092] That is, if the absolute value of the second trend parameter c d is greater than or equal to the first threshold value V1, and the absolute value of the third trend parameter c s is greater than the second threshold value V2, and the signs of the second trend parameter c d and the third trend parameter c s are opposite, and the second trend parameter c d is greater than zero, then the number of the target torque fitting line segment is a fifth number 3, and the fifth number represents a trend of the target torque fitting line segment being suddenly increased compared with the previous torque fitting line segment.

[0093] Step 8, if the absolute value of the second trend parameter is greater than or equal to the first threshold value, and the absolute value of the third trend parameter is greater than the second threshold value, and the signs of the second trend parameter and the third trend parameter are opposite, and the second trend parameter is less than zero, then the number of the target torque fitting line segment is a sixth number.

[0094] That is, if the absolute value of the second trend parameter c d is greater than or equal to the first threshold value V1, and the absolute value of the third trend parameter c s is greater than the second threshold value V2, and the signs of the second trend parameter c d and the third trend parameter c s are opposite, and the second trend parameter c d is less than zero, then the number of the target torque fitting line segment is a sixth number -3, and the sixth number represents a trend of the target torque fitting line segment being suddenly decreased compared with the previous torque fitting line segment.

[0095] Step 9, determining the training torque matrix according to the numbers corresponding to the respective torque fitting line segments.

[0096] It should be noted that in other embodiments of the present application, when the relationship between the first trend parameter, the second trend parameter and the third trend parameter of the torque fitting line segment and the preset first threshold value and the second threshold value does not belong to any of the above cases, then the number of the target torque fitting line segment is a seventh number 0, wherein the seventh number represents a trend of the target torque fitting line segment being stable compared with the previous torque fitting line segment.

[0097] It should be noted that, as shown in FIG. 1, a schematic diagram of stable, up-step, down-step, up, down, sudden increase and sudden decrease is shown. Figure 5

[0098] ​In other embodiments of the present application, the method further comprises: obtaining normal training torque data and corresponding multiple abnormal training torque data, specifically:

[0099] Step 1, obtaining an angle displacement group.

[0100] The angle displacement group is pre-stored historical data, including multiple end segment angular displacements, i.e., the angular displacement of the end segment joint of the robot arm, which lays a foundation for subsequent calculation of normal training torque data and corresponding multiple abnormal training torque data.

[0101] Step 2, based on a preset angular velocity calculation formula and an angular acceleration calculation formula, determining an angular velocity group and an angular acceleration group according to the angle displacement group.

[0102] The multiple angular velocity values in the angular velocity group and the multiple angular acceleration values in the angular acceleration group correspond to the multiple end segment angular displacements in the angle displacement group.

[0103] Exemplarily, the angular velocity calculation formula is: .

[0104] Exemplarily, the angular acceleration calculation formula is: .

[0105] That is, each end segment angular displacement in the angle displacement group is substituted into the preset angular velocity calculation formula and angular acceleration calculation formula for calculation, thereby determining the angular velocity group and the angular acceleration.

[0106] Step 3, based on a preset torque calculation formula, determining normal training torque data according to the angle displacement group, the angular velocity group and the angular acceleration group.

[0107] That is, the end segment angular displacement in the angle displacement group, the corresponding angular velocity and the corresponding angular acceleration are substituted into the preset torque calculation formula, thereby determining the normal training torque data.

[0108] Exemplarily, the torque calculation formula is:

[0109] , wherein, represents the angular displacement of the end segment joint of the robot arm, i.e., the angle of rotation of the object around the shaft, is the angular velocity of the end segment joint of the robot arm, i.e., the rate of change of the angular displacement with time, and the unit is rad / s, and is the angular acceleration of the end segment joint of the robot arm, i.e. the rate of change with time, M(.) represents the robot arm mass moment function, which represents the inertia property of the combination of the robot arm mass and acceleration; C(.) represents the centrifugal force function of the robot arm, which represents the centrifugal force of the robot arm at a certain angular velocity; G(.) represents the gravity function, i.e., the torque generated by the gravity (earth's gravity) on each joint.

[0110] It should be noted that the torque calculation formula can be used to construct a mechanical arm simulation model as shown in Figure 6 The simulation model can convert the input mechanical arm running angular displacement into torque simulation data.

[0111] Step 4, based on the preset first abnormal formula, the first abnormal training torque data is determined according to the angular displacement group, the angular velocity group and the angular acceleration group.

[0112] It should be noted that the first abnormal formula corresponds to the abnormal formula when the mechanical arm joint appears to be stuck, that is, a random constant is added to the torque calculation formula to make the result relative to the normal training torque data. Step.

[0113] That is, the last angular displacement in the angular displacement group, the corresponding angular velocity and the corresponding angular acceleration are substituted into the preset first abnormal formula, so as to determine the first abnormal training torque data.

[0114] Step 5, based on the preset second abnormal formula, the second abnormal training torque data is determined according to the angular displacement group, the angular velocity group and the angular acceleration group.

[0115] It should be noted that the second abnormal formula corresponds to the abnormal formula when the mechanical arm joint appears to be free, that is, a short-time random torque constant is subtracted from the torque calculation formula to make the result relative to the normal training torque data. Step.

[0116] That is, the last angular displacement in the angular displacement group, the corresponding angular velocity and the corresponding angular acceleration are substituted into the preset second abnormal formula, so as to determine the second abnormal training torque data.

[0117] Step 6, based on the preset third abnormal formula, the third abnormal training torque data is determined according to the angular displacement group, the angular velocity group and the angular acceleration group.

[0118] It should be noted that the third abnormal formula corresponds to the abnormal formula when the mechanical arm joint appears to be aged, that is, a linear rising function is added to the torque calculation formula to make it relative to the normal training torque data. Step.

[0119] That is, the last angular displacement in the angular displacement group, the corresponding angular velocity and the corresponding angular acceleration are substituted into the preset third abnormal formula, so as to determine the third abnormal training torque data.

[0120] Step 7, according to the first abnormal training torque data, the second abnormal training torque data and the third abnormal training torque data, the abnormal training torque data is determined.

[0121] That is, the first abnormal training torque data, the second abnormal training torque data, and the third abnormal training torque data are taken as the abnormal training torque data.

[0122] It should be noted that in other embodiments of the present application, the preset database further includes normal training position data, i.e., the position of the last arm rod of the robot arm.

[0123] Exemplarily, the normal training position data can be determined based on a robot arm positioning formula, where the robot arm positioning formula is:

[0124]

[0125] wherein, is the position of the last arm rod of the robot arm relative to the position of the Cartesian coordinate system of joint A. is a transformation matrix of the Cartesian coordinate system of joint B to the Cartesian coordinate system of joint A, is the position of the last arm rod of the robot arm relative to the position of the Cartesian coordinate system of joint B. Wherein, the and are determined. .

[0126] It should be noted that the chain rule can be used to obtain the derivation formula of other arbitrary extended form coordinate system 0 to arbitrary coordinate system n, specifically:

[0127]

[0128] wherein, is a transformation matrix of the Cartesian coordinate system of joint n to the Cartesian coordinate system of joint 0, is a transformation matrix of the Cartesian coordinate system of joint 1 to the Cartesian coordinate system of joint 0, is a transformation matrix of the Cartesian coordinate system of joint 2 to the Cartesian coordinate system of joint 1, is a transformation matrix of the Cartesian coordinate system of joint n to the Cartesian coordinate system of joint n-1.

[0129] It should be noted that, Specifically,

[0130] wherein, θ n represents the included angle between the x-axes of the two Cartesian coordinate systems at joint n and joint n-1, θ n is the angular displacement of the last joint of the robot arm as described above , represents the included angle between the z-axes of the two Cartesian coordinate systems at joint n and joint n-1, and a ndenotes the distance between the z-axes of the two Cartesian coordinate systems at joint n and joint n-1, and is a constant value, d n denotes the distance between the x-axes of the two Cartesian coordinate systems at joint n and joint n-1, and is a constant value. The corresponding position information of the coordinate system 0 is obtained by substituting the different coordinate system θ into T0.

[0131] In other embodiments of the present application, the actual difference data set of the robot arm is obtained, and the actual torque matrix is determined by:

[0132] Step 1, obtaining a plurality of instruction sets, each instruction set comprising a plurality of angular displacement instructions.

[0133] The angular displacement instructions are used to control the corresponding joint of the robot arm to move, i.e., to control the corresponding joint of the robot arm to make a corresponding angular displacement.

[0134] It should be noted that one instruction set comprises a plurality of angular displacement instructions, each angular displacement instruction controls the corresponding joint of the robot arm to make a corresponding angular displacement, thereby controlling the robot arm to perform the operation corresponding to the instruction set.

[0135] Step 2, controlling the robot arm according to the plurality of instruction sets respectively, and obtaining the actual torque of the robot arm corresponding to each instruction set to determine the actual torque set.

[0136] That is, the robot arm is controlled to perform the operation corresponding to the instruction set by the plurality of instruction sets respectively, and the actual torque of the robot arm is sampled by the sensor, thereby determining the actual torque set. The actual torque is the torque of the end joint of the robot arm.

[0137] Step 3, determining the ideal torque set according to the plurality of instruction sets based on the preset angular velocity calculation formula, angular acceleration calculation formula and torque calculation formula.

[0138] That is, the angular displacement in the angular displacement instruction corresponding to the end joint of the robot arm is taken as the ideal end angular displacement, and each ideal end angular displacement is substituted into the angular velocity calculation formula and the angular acceleration calculation formula, thereby determining a plurality of ideal angular velocities and ideal angular accelerations. The ideal end angular displacement, the corresponding ideal angular velocity and the corresponding ideal angular acceleration are substituted into the torque calculation formula, thereby determining the ideal torque set.

[0139] The plurality of ideal torques in the ideal torque set correspond one-to-one to the plurality of actual torques in the actual torque set.

[0140] Step 4, determining the actual difference data set by subtracting the actual torque set from the ideal torque set.

[0141] That is, the ideal torque in the ideal torque set is subtracted from the actual torque in the corresponding actual torque set respectively, thereby determining the actual difference data set.

[0142] That is, the number of ideal torques in the ideal torque group is the same as the number of actual torques in the actual torque group, and there is a one-to-one correspondence relationship, specifically, a plurality of residual values are obtained by respectively subtracting the ideal torque in the ideal torque group from the corresponding actual torque in the actual torque group, as the actual difference value data group.

[0143] Step 5, linear fitting is performed according to the actual difference value data group to determine a plurality of fitting line segments.

[0144] That is, linear fitting is performed on the actual difference value data group so that the distance between a plurality of actual difference value data in the actual difference value data group and the corresponding fitting line segment is within a preset range, and the plurality of fitting line segments. This step is the same as the step of performing linear fitting on the target torque residual sequence corresponding to the target abnormal training torque data in the above-mentioned embodiment to determine a plurality of torque fitting line segments, which will not be described here.

[0145] Step 6, determining the actual torque matrix according to the preset trend coding rule and the plurality of fitting line segments.

[0146] That is, the first trend parameter, the second trend parameter and the third trend parameter of each fitting line segment are calculated, and the actual torque matrix is finally determined based on the preset trend coding rule. This step is the same as the step of determining the training torque matrix in the above-mentioned embodiment, which will not be described here.

[0147] The actual torque matrix is used to represent the actual state of the robot arm, and the fault state of the robot arm can be determined by the target fault diagnosis model according to the actual torque matrix for fault prediction.

[0148] Exemplarily, as shown in Figure 7 , a system architecture diagram of a robot arm fault diagnosis method is shown, a feature extraction method of generating a high-dimensional qualitative trend primitive matrix based on residual values based on non-contact measuring instrument marking results, and a robot arm fault diagnosis method based on deep learning. The robot arm fault diagnosis method collects data information such as motor torque, speed and displacement of each joint of the robot arm through the PLC. In order to improve the quality of modeling data, the time stamp sequence of the corresponding data needs to be collected, and after the robot arm is installed each time, the installation tolerance of the robot arm is measured by the non-contact measuring instrument as the non-contact measuring instrument marking result and the label is generated, thereby serving as the basis for performance evaluation of the target fault diagnosis model. The collected data is input into the industrial computer for fault diagnosis, and the related process data and result data are stored in the memory and displayed through the display.

[0149] In a specific embodiment of the present application, as Figure 8As shown, a flow architecture diagram of the mechanical arm fault diagnosis method is shown, and the mechanical arm fault diagnosis method includes modeling, monitoring and diagnosis three parts. In the modeling stage, after the angular displacement group is pre-acquired, the angular velocity group and the angular acceleration group are determined according to the angular velocity calculation formula and the angular acceleration calculation formula, the normal simulation is realized based on the preset simulation model, that is, based on the torque calculation formula, the normal training torque data is determined, then the fault simulation is realized through the fault mechanism injection, so as to determine the abnormal training torque data, then the residual calculation is performed, and the corresponding label is generated according to the type of abnormal state. The result of residual calculation, and finally generate the qualitative matrix, that is, the training torque matrix, to train the preset initial fault diagnosis model and determine the target fault diagnosis model.

[0150] In the monitoring stage, a plurality of instruction groups are acquired, and the ideal torque group is determined based on the observer, that is, the preset torque calculation formula. The controlled device, that is, the mechanical arm, executes the operation corresponding to the instruction group based on the actuator control, and the torque during the operation is sampled by the sensor, so as to obtain the actual torque group. The residual calculation is performed according to the actual torque group and the ideal torque group to determine the actual torque matrix.

[0151] In the diagnosis stage, the actual torque matrix is input into the deep learning model, that is, the target fault diagnosis model, for fault prediction, so as to determine the fault state of the mechanical arm.

[0152] It should be noted that the mechanical arm fault diagnosis method constructs the running simulation model of the mechanical arm in the virtual space, thereby obtaining rich abnormal training torque data and normal training torque data, solving the problem of missing model training data. In addition, when the model needs to be upgraded, only the new abnormal training torque data needs to be input to quickly complete the deep learning model training and upgrading.

[0153] And the normal training torque data and the corresponding abnormal training torque data are obtained through normal simulation and fault simulation, respectively, the theoretical position of the mechanical arm end is calculated through the observer, and the actual torque and the ideal torque are compared and the residual is calculated, then the high-dimensional qualitative trend primitive matrix, that is, the training torque matrix and the actual torque matrix, are further generated according to the residual. The training torque matrix and the actual torque matrix are represented in the physical dimension for multiple correlation measurement points in the row, and represented in the time dimension for the qualitative trend of each measurement point in the column. The characteristics of the high-dimensional convolution kernel are used to simultaneously obtain the physical and time dimensions of the input matrix for high-dimensional feature extraction and fault classification. On the basis of retaining high-dimensional information, a large amount of irrelevant information can be eliminated to avoid the occurrence of the situation that the fault diagnosis model is difficult to converge due to data redundancy, reduce the amount of data required for model training, and increase the speed of training model and the robustness of model.

[0154] According to the actual situation, the surface mounting tolerance of the mechanical arm installation result can be measured remotely by a non-contact measuring instrument, and the inference result of the fault diagnosis model is further evaluated according to the surface mounting tolerance, and the installation criterion result is given as a label for quality verification of the fault diagnosis model, so as to further improve the performance of the model.

[0155] In other embodiments of the present application, when the surface mounting tolerance is large and the inference result of the mechanical arm fault diagnosis model is normal, or the installation tolerance is within the normal range but the inference result of the mechanical arm fault diagnosis model is that the mechanical arm fails, it is considered that the inference model may be abnormal, and the model needs to be retrained.

[0156] The above mechanical arm fault diagnosis method generates normal training torque data and abnormal training torque data for model training according to mechanical structure and dynamics to construct a mechanical arm operation simulation model, and generates sufficient samples for training and constructing the fault diagnosis model through fault mechanism injection. The model input is a high-dimensional qualitative trend primitive matrix generated based on fault and abnormal data residual values. The matrix is a multi-correlation measuring point in the physical dimension in the row direction and a qualitative trend change of each point in the time dimension in the column direction. The matrix is used as the input of the model, which can effectively eliminate single-dimensional redundant information and retain high-dimensional effective information to improve the training speed and robustness of the model, so that the constructed model can quickly and accurately identify the health status of the mechanical arm. The fault diagnosis model itself can measure the surface tolerance of the installed components of the mechanical arm by a non-contact measuring instrument to judge and give the evaluation result of the reliability of the fault inference model.

[0157] It should be understood that, although each step in the flowchart involved in each embodiment as described above is displayed in sequence according to the arrow, these steps are not necessarily executed in the order indicated by the arrow. Unless otherwise specified herein, the execution of these steps is not strictly limited in sequence, and these steps can be executed in other orders. Moreover, at least part of the steps in the flowchart involved in each embodiment as described above can include multiple steps or stages, which are not necessarily executed at the same time, but can be executed at different times, and the execution sequence of these steps or stages is not necessarily sequential, but can be executed in rotation or alternation with at least part of other steps or steps or stages in other steps.

[0158] Based on the same inventive concept, the embodiments of the present application further provide a robot arm fault diagnosis device for implementing the robot arm fault diagnosis method described above. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme described in the above method, so the specific limitations in one or more robot arm fault diagnosis device embodiments provided below can refer to the limitations of the robot arm fault diagnosis method described above, which will not be repeated here.

[0159] In one embodiment of the present application, as shown in Figure 9 A robot arm fault diagnosis device is provided, comprising:

[0160] The construction module 100 is configured to construct a training torque matrix based on normal training torque data and a plurality of abnormal training torque data in a preset database.

[0161] The training module 200 is configured to train a preset initial fault diagnosis model according to the training torque matrix, and determine a target fault diagnosis model.

[0162] The acquisition module 300 is configured to acquire actual difference data sets of the robot arm, and determine an actual torque matrix.

[0163] The determination module 400 is configured to perform fault prediction according to the actual torque matrix based on the target fault diagnosis model, and determine a fault state of the robot arm.

[0164] In an embodiment of the present application, the construction module 100 is further configured to obtain a torque residual sequence corresponding to each abnormal training torque data by subtracting the normal training torque data from the abnormal training torque data of each fault type; determine the training torque matrix according to the torque residual sequence corresponding to each abnormal training torque data; and each torque residual sequence in the training torque matrix corresponds to a fault type label.

[0165] In an embodiment of the present application, the construction module 100 is further configured to perform linear fitting on a target torque residual sequence corresponding to target abnormal training torque data to determine a plurality of torque fitting line segments; the target abnormal training torque data is any abnormal training torque data in the plurality of abnormal training torque data; and determine the training torque matrix according to a preset trend coding rule and the plurality of torque fitting line segments of each abnormal training torque data.

[0166] In some embodiments of the present application, the construction module 100 is further configured to determine a first trend parameter of the target torque fitting line segment according to the last residual value of the target torque fitting line segment and the last residual value of the previous torque fitting line segment of the target torque fitting line segment; determine a second trend parameter of the target torque fitting line segment according to the first residual value of the target torque fitting line segment and the last residual value of the previous torque fitting line segment of the target torque fitting line segment; determine a third trend parameter of the target torque fitting line segment according to the first residual value of the target torque fitting line segment and the last residual value of the target torque fitting line segment; and determine the training torque matrix according to the first trend parameter, the second trend parameter and the third trend parameter based on the first threshold value and the second threshold value.

[0167] In some embodiments of the present application, the constructing module 100 is further configured to: if the absolute value of the second trend parameter is less than the first threshold value, and the absolute value of the first trend parameter is greater than the second threshold value, and the first trend parameter is greater than zero, then the number of the target torque fitting line segment is a first number; if the absolute value of the second trend parameter is less than the first threshold value, and the absolute value of the first trend parameter is greater than or equal to the second threshold value, and the first trend parameter is less than zero, then the number of the target torque fitting line segment is a second number; if the absolute value of the second trend parameter is greater than or equal to the first threshold value, and the absolute value of the third trend parameter is less than or equal to the second threshold value, and the first trend parameter is greater than zero, then the number of the target torque fitting line segment is a third number; if the absolute value of the second trend parameter is greater than or equal to the first threshold value, and the absolute value of the third trend parameter is less than or equal to the second threshold value, and the first trend parameter is less than zero, then the number of the target torque fitting line segment is a fourth number; if the absolute value of the second trend parameter is greater than or equal to the first threshold value, and the absolute value of the third trend parameter is greater than the second threshold value, and the signs of the second trend parameter and the third trend parameter are the same, and the second trend parameter is greater than zero, then the number of the target torque fitting line segment is the first number; if the absolute value of the second trend parameter is greater than or equal to the first threshold value, and the absolute value of the third trend parameter is greater than the second threshold value, and the signs of the second trend parameter and the third trend parameter are the same, and the second trend parameter is less than zero, then the number of the target torque fitting line segment is the second number; if the absolute value of the second trend parameter is greater than or equal to the first threshold value, and the absolute value of the third trend parameter is greater than the second threshold value, and the signs of the second trend parameter and the third trend parameter are opposite, and the second trend parameter is greater than zero, then the number of the target torque fitting line segment is a fifth number; if the absolute value of the second trend parameter is greater than or equal to the first threshold value, and the absolute value of the third trend parameter is greater than the second threshold value, and the signs of the second trend parameter and the third trend parameter are opposite, and the second trend parameter is less than zero, then the number of the target torque fitting line segment is a sixth number; and determining the training torque matrix according to the numbers corresponding to the respective torque fitting line segments.

[0168] In an embodiment of the present application, the acquisition module 300 is further configured to acquire a group of angular displacements; the angular displacements in the group of angular displacements are end segment angular displacements of the robot arm; based on a preset angular velocity calculation formula and an angular acceleration calculation formula, the group of angular displacements is used to determine a group of angular velocities and a group of angular accelerations; a plurality of angular velocity values in the group of angular velocities and a plurality of angular acceleration values in the group of angular accelerations correspond to a plurality of end segment angular displacements in the group of angular displacements; based on a preset torque calculation formula, the group of angular displacements, the group of angular velocities, and the group of angular accelerations are used to determine normal training torque data; based on a preset first abnormal formula, the group of angular displacements, the group of angular velocities, and the group of angular accelerations are used to determine first abnormal training torque data; based on a preset second abnormal formula, the group of angular displacements, the group of angular velocities, and the group of angular accelerations are used to determine second abnormal training torque data; based on a preset third abnormal formula, the group of angular displacements, the group of angular velocities, and the group of angular accelerations are used to determine third abnormal training torque data; and the first abnormal training torque data, the second abnormal training torque data, and the third abnormal training torque data are used to determine abnormal training torque data.

[0169] In an embodiment of the present application, the acquisition module 300 is further configured to acquire a plurality of groups of instructions, the group of instructions including a plurality of angular displacement instructions; the angular displacement instructions are used to control corresponding joint movements of the robot arm; the robot arm is controlled according to the plurality of groups of instructions, and actual torque of the robot arm corresponding to each of the groups of instructions is acquired to determine an actual torque group; based on a preset angular velocity calculation formula, an angular acceleration calculation formula, and a torque calculation formula, the plurality of groups of instructions are used to determine an ideal torque group; a plurality of ideal torques in the ideal torque group correspond to a plurality of actual torques in the actual torque group in a one-to-one manner; the ideal torque group and the actual torque group are subtracted to determine an actual difference value group; the actual difference value group is linearly fitted to determine a plurality of fitting line segments; and the actual torque matrix is determined according to a preset trend coding rule and the plurality of fitting line segments.

[0170] The above-mentioned various modules in the robot arm fault diagnosis apparatus can be realized by software, hardware, or a combination thereof, in whole or in part. The above-mentioned various modules can be embedded in or independent of a processor in a computer device in a hardware form, or can be stored in a memory in a computer device in a software form, so as to be called and executed by a processor to perform operations corresponding to the above-mentioned various modules.

[0171] In an embodiment of the present application, a computer device is provided, which can be a server, and an internal structure diagram of the computer device can be as shown in FIG. 8. Figure 10As shown in the figure. The computer device includes a processor, a memory and a network interface connected by a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operating system and the computer program in the non-volatile storage medium to run. The database of the computer device is used to store all related data for executing the robot fault diagnosis method. The network interface of the computer device is used to communicate with the external terminal through the network connection. The computer program is executed by the processor to implement a robot fault diagnosis method.

[0172] Those skilled in the art can understand that, Figure 10 The structure shown in the figure is only a block diagram of part of the structure related to the scheme of the present application, and does not constitute a limitation on the computer device to which the scheme of the present application is applied. The specific computer device can include more or fewer components than those shown in the figure, or combine certain components, or have a different component arrangement.

[0173] In an embodiment of the present application, a computer device is provided, including a memory and a processor, the memory stores a computer program, and the processor executes the computer program to implement the steps of the robot fault diagnosis method in the above-mentioned embodiments.

[0174] In an embodiment of the present application, a computer readable storage medium is provided, which stores a computer program, and the computer program is executed by a processor to implement the steps of the robot fault diagnosis method in the above-mentioned method embodiments.

[0175] In an embodiment of the present application, a computer program product is provided, including a computer program, and the computer program is executed by a processor to implement the steps of the robot fault diagnosis method in the above-mentioned method embodiments.

[0176] It should be noted that the user information (including but not limited to user equipment information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) involved in the present application are all information and data authorized by the user or authorized by all parties.

[0177] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0178] The technical features of the above embodiments can be combined in any way. For the sake of brevity, 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.

[0179] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A robot arm failure diagnosis method characterized by comprising: The method comprises: Based on the preset database, the normal training torque data and the corresponding multiple abnormal training torque data are used to construct a training torque matrix; According to the training torque matrix, the initial fault diagnosis model is trained to determine the target fault diagnosis model; Obtain the actual difference data set of the mechanical arm, and determine the actual torque matrix; Based on the target fault diagnosis model, the fault prediction is performed according to the actual torque matrix to determine the fault state of the mechanical arm.

2. The robot arm failure diagnosis method according to claim 1, characterized in that, The multiple abnormal training torque data comprises: abnormal training torque data corresponding to multiple fault types; The timestamps of the normal training torque data correspond to the timestamps of each abnormal training torque data; The training torque matrix is constructed based on the normal training torque data and the corresponding multiple abnormal training torque data in the preset database, comprising: The normal training torque data is subtracted from each fault type abnormal training torque data to obtain a torque residual sequence corresponding to each abnormal training torque data; According to the torque residual sequence corresponding to each abnormal training torque data, the training torque matrix is determined; Each torque residual sequence in the training torque matrix corresponds to a fault type label.

3. The method of claim 2, wherein According to the torque residual sequence corresponding to each abnormal training torque data, the training torque matrix is determined, comprising: According to the target torque residual sequence corresponding to the target abnormal training torque data, a plurality of torque fitting line segments are determined; The target abnormal training torque data is any abnormal training torque data in the multiple abnormal training torque data; According to the preset trend coding rule and the plurality of torque fitting line segments of each abnormal training torque data, the training torque matrix is determined.

4. The robot arm failure diagnosis method according to claim 3, characterized by, According to the preset trend coding rule and the plurality of torque fitting line segments of each abnormal training torque data, the training torque matrix is determined, comprising: According to the last residual value of the target torque fitting line segment and the last residual value of the previous torque fitting line segment of the target torque fitting line segment, the first trend parameter of the target torque fitting line segment is determined; According to the first residual value of the target torque fitting line segment and the last residual value of the previous torque fitting line segment of the target torque fitting line segment, the second trend parameter of the target torque fitting line segment is determined; According to the first residual value of the target torque fitting line segment and the last residual value of the target torque fitting line segment, the third trend parameter of the target torque fitting line segment is determined; Based on the preset first threshold and second threshold, the first trend parameter, the second trend parameter and the third trend parameter are used to determine the training torque matrix.

5. The method of claim 4, wherein Based on the preset first threshold and second threshold, the first trend parameter, the second trend parameter and the third trend parameter are used to determine the training torque matrix, comprising: Based on the preset first threshold and second threshold, the first trend parameter, the second trend parameter and the third trend parameter are used to determine the number of the target torque fitting line segment as the first number or the second number; Based on the preset first threshold and second threshold, the second trend parameter and the third trend parameter are used to determine the number of the target torque fitting line segment as the third number or the fourth number; Determine the number of the target torque fitting line segment as the fifth number or the sixth number according to the second trend parameter and the third trend parameter based on the preset first threshold value and the second threshold value; Determine the training torque matrix according to the number corresponding to each torque fitting line segment.

6. The robot arm failure diagnosis method according to claim 1, characterized by, The method further comprises: Obtain an angular displacement group; the angular displacement in the angular displacement group is the end segment angular displacement of the mechanical arm; Determine an angular velocity group and an angular acceleration group according to the angular displacement group based on a preset angular velocity calculation formula and an angular acceleration calculation formula; the multiple angular velocity values in the angular velocity group and the multiple angular acceleration values in the angular acceleration group correspond to the multiple end segment angular displacements in the angular displacement group; Determine normal training torque data according to the angular displacement group, the angular velocity group and the angular acceleration group based on a preset torque calculation formula; Determine first abnormal training torque data according to the angular displacement group, the angular velocity group and the angular acceleration group based on a preset first abnormal formula; Determine second abnormal training torque data according to the angular displacement group, the angular velocity group and the angular acceleration group based on a preset second abnormal formula; Determine third abnormal training torque data according to the angular displacement group, the angular velocity group and the angular acceleration group based on a preset third abnormal formula; Determine abnormal training torque data according to the first abnormal training torque data, the second abnormal training torque data and the third abnormal training torque data.

7. The robot arm failure diagnosis method according to claim 1, characterized by, The obtaining of the actual difference value data group of the mechanical arm and the determination of the actual torque matrix comprise: Obtain multiple instruction groups, the instruction group comprising multiple angular displacement instructions; the angular displacement instruction being used for controlling the corresponding joint movement of the mechanical arm; Control the mechanical arm according to multiple instruction groups respectively, and obtain the actual torque of the mechanical arm corresponding to each instruction group to determine an actual torque group; Determine an ideal torque group according to multiple instruction groups based on a preset angular velocity calculation formula, an angular acceleration calculation formula and a torque calculation formula; the multiple ideal torques in the ideal torque group corresponding to the multiple actual torques in the actual torque group one by one; Determine an actual difference value data group by subtracting the ideal torque group from the actual torque group; Determine multiple fitting line segments by linear fitting according to the actual difference value data group; Determine the actual torque matrix according to a preset trend coding rule and the multiple fitting line segments.

8. A robot arm failure diagnosis device characterized by comprising: The device comprises: A construction module configured to construct a training torque matrix based on normal training torque data and corresponding multiple abnormal training torque data in a preset database; A training module configured to train a preset initial fault diagnosis model according to the training torque matrix to determine a target fault diagnosis model; An acquisition module configured to obtain an actual difference value data group of a mechanical arm and determine an actual torque matrix; A determination module configured to perform fault prediction according to the actual torque matrix based on the target fault diagnosis model to determine a fault state of the mechanical arm. 9.A computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the computer device is configured to perform the method according to any one of claims 1-8 when the computer program is executed by the processor. The processor executes the computer program to implement the steps of the method in any one of claims 1 to 7.

10. A computer-readable storage medium having stored thereon a computer program, characterized in that, The computer program is executed by the processor to implement the steps of the method in any one of claims 1 to 7.

Images