Ball screw fault detection method and device and computer equipment

Abstract

The invention relates to a ball screw fault detection method and device and computer equipment. The method comprises the steps that the real-time motor torque of a servo motor connected with a target ball screw is collected, and a real-time torque order sequence is determined according to the real-time motor torque; according to the real-time screw rotation speed, the real-time screw nominal diameter, the real-time ball nominal diameter, the real-time screw thread lead angle and the real-time contact angle of the target ball screw, the real-time screw meshing frequency, the real-time nut meshing frequency and the real-time rolling body meshing frequency are determined; and inputting the real-time torque order sequence, the real-time screw meshing frequency, the real-time nut meshing frequency and the real-time rolling body meshing frequency into a pre-trained fault detection model, and determining fault information of the target ball screw. According to the scheme, the fault detection efficiency of the ball screw can be improved, the fault detection precision of the ball screw is improved, and the fault detection cost of the ball screw is reduced.

Description

Technical Field

[0001] This application relates to the field of equipment testing technology, and in particular to a method, apparatus and computer equipment for fault detection of ball screws. Background Technology

[0002] Ball screws convert rotary motion into precise linear motion and are widely used in CNC machine tools, automated equipment, precision machinery, and the nuclear industry, bearing the crucial responsibility of transmitting power and achieving positioning. However, during long-term operation, ball screws may experience abnormalities due to factors such as wear, insufficient lubrication, overload, and manufacturing defects. These abnormalities can manifest as vibration, noise, and temperature rise, potentially leading to equipment failure and impacting production efficiency and accuracy. Currently, abnormal condition monitoring methods using vibration sensors, temperature sensors, current sensors, or visual inspection are commonly employed. While these methods can detect some anomalies, they often require complex hardware and installation space, resulting in high costs for ball screw fault detection and failing to provide timely and accurate information about the ball screw's operating status. Therefore, improving the efficiency and accuracy of ball screw fault detection while reducing its costs is a problem that needs to be addressed. Summary of the Invention

[0003] Therefore, it is necessary to provide a method, apparatus, and computer equipment for detecting ball screw faults that can improve the efficiency and accuracy of ball screw fault detection, in order to address the aforementioned technical problems.

[0004] In a first aspect, this application provides a method for detecting faults in a ball screw, the method comprising:

[0005] The real-time motor torque of the servo motor connected to the target ball screw is collected, and the real-time torque order sequence is determined based on the real-time motor torque.

[0006] The real-time screw engagement frequency, real-time nut engagement frequency, and real-time rolling element engagement frequency are determined based on the real-time screw rotation speed, real-time screw nominal diameter, real-time ball nominal diameter, real-time screw thread helix angle, and real-time contact angle of the target ball screw.

[0007] The real-time torque order sequence, the real-time ball screw engagement frequency, the real-time nut engagement frequency, and the real-time rolling element engagement frequency are input into a pre-trained fault detection model to determine the fault information of the target ball screw.

[0008] In one embodiment, the real-time torque order sequence, the real-time ball screw engagement frequency, the real-time nut engagement frequency, and the real-time rolling element engagement frequency are input into a pre-trained fault detection model to determine the fault information of the target ball screw, including:

[0009] The real-time torque order sequence, the real-time lead screw meshing frequency, the real-time nut meshing frequency, and the real-time rolling element meshing frequency are input into a pre-trained fault detection model so that the fault detection model can determine the convolution kernel operator based on the real-time torque order sequence, the real-time lead screw meshing frequency, the real-time nut meshing frequency, and the real-time rolling element meshing frequency.

[0010] The real-time fault detection index is determined by the convolutional layer of the fault detection model based on the convolutional kernel operator, the real-time lead screw meshing frequency, the real-time nut meshing frequency, and the real-time rolling element meshing frequency; the real-time fault detection index includes: real-time lead screw meshing order, real-time nut meshing order, and real-time rolling element meshing order.

[0011] The target fault perception features are determined based on the real-time fault detection index through the attention network layer of the fault detection model.

[0012] The fault information of the target ball screw is determined by the multilayer sensor of the fault detection model based on the target fault perception characteristics.

[0013] In one embodiment, the target fault perception features are determined based on the real-time fault detection metrics through the attention network layer of the fault detection model, including:

[0014] The target input matrix is ​​determined based on the real-time fault detection index, and the query vector matrix, key vector matrix, and value vector matrix are determined based on the target input matrix through the attention network layer of the fault detection model.

[0015] The attention matrix is ​​determined based on the query vector matrix and the key vector matrix, and the intermediate representation vector is determined based on the attention matrix and the value vector matrix.

[0016] The real-time fault detection indicators are weighted based on the intermediate representation vector to determine the fault perception features.

[0017] In one embodiment, the fault information of the target ball screw is determined based on the target fault detection features using a multilayer perceptron of the fault detection model, including:

[0018] The multilayer perceptron of the fault detection model processes the target fault perception features based on a normalized activation function to determine the probability corresponding to the fault category of the target ball screw; the fault category includes screw fault, nut fault, rolling element fault, and no fault.

[0019] The fault information of the target ball screw is determined based on the probability corresponding to the fault category.

[0020] In one embodiment, the real-time screw engagement frequency, real-time nut engagement frequency, and real-time rolling element engagement frequency are determined based on the real-time screw rotation speed, real-time screw nominal diameter, real-time ball nominal diameter, real-time screw thread helix angle, and real-time contact angle of the target ball screw, including:

[0021] The real-time lead screw engagement frequency is determined based on the real-time lead screw rotation speed, real-time lead screw nominal diameter, real-time ball nominal diameter, real-time lead screw thread helix angle, and real-time contact angle.

[0022] The real-time nut engagement frequency is determined based on the real-time screw rotation speed, real-time screw nominal diameter, real-time ball nominal diameter, real-time screw thread helix angle, and real-time contact angle.

[0023] The real-time rolling element engagement frequency is determined based on the real-time lead screw rotation speed, real-time lead screw nominal diameter, real-time ball nominal diameter, and real-time contact angle.

[0024] In one embodiment, the above-mentioned fault detection method for ball screws further includes:

[0025] The sample motor torque of the servo motor connected to the sample ball screw is determined, the sample torque order sequence is determined based on the sample motor torque, and the sample state information of the sample ball screw is determined; the sample state information includes screw fault state, nut fault state, rolling element fault state, and normal state.

[0026] The neural network model is trained based on the sample torque order sequence, sample lead screw meshing frequency, sample nut meshing frequency, sample rolling element meshing frequency, and the sample state information to determine the fault detection model.

[0027] In one embodiment, a neural network model is trained based on the sample torque order sequence, sample lead screw meshing frequency, sample nut meshing frequency, sample rolling element meshing frequency, and the sample state information to determine a fault detection model, including:

[0028] The neural network model is trained based on the sample torque order sequence, sample lead screw meshing frequency, sample nut meshing frequency, sample rolling element meshing frequency, and the sample state information, and the cross-entropy loss function of the neural network model is determined in real time during the training process.

[0029] The gradient descent method is used to update the model parameters of the neural network model based on the cross-entropy loss function until the neural network model converges. The trained neural network model is then used as the fault detection model.

[0030] Secondly, this application also provides a fault detection device for ball screws, the device comprising:

[0031] The torque order sequence determination module is used to collect the real-time motor torque of the servo motor connected to the target ball screw, and determine the real-time torque order sequence based on the real-time motor torque.

[0032] The meshing frequency determination module is used to determine the real-time screw meshing frequency, real-time nut meshing frequency, and real-time rolling element meshing frequency based on the real-time screw rotation speed, real-time screw nominal diameter, real-time ball nominal diameter, real-time screw thread helix angle, and real-time contact angle of the target ball screw.

[0033] The fault information determination module is used to input the real-time torque order sequence, the real-time ball screw engagement frequency, the real-time nut engagement frequency, and the real-time rolling element engagement frequency into a pre-trained fault detection model to determine the fault information of the target ball screw.

[0034] Thirdly, this application also provides a computer device, the computer device including a memory and a processor, the memory storing a computer program, and the processor executing the computer program to perform the following steps:

[0035] The real-time motor torque of the servo motor connected to the target ball screw is collected, and the real-time torque order sequence is determined based on the real-time motor torque.

[0036] The real-time screw engagement frequency, real-time nut engagement frequency, and real-time rolling element engagement frequency are determined based on the real-time screw rotation speed, real-time screw nominal diameter, real-time ball nominal diameter, real-time screw thread helix angle, and real-time contact angle of the target ball screw.

[0037] The real-time torque order sequence, the real-time ball screw engagement frequency, the real-time nut engagement frequency, and the real-time rolling element engagement frequency are input into a pre-trained fault detection model to determine the fault information of the target ball screw.

[0038] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, the computer program performing the following steps when executed by a processor:

[0039] The real-time motor torque of the servo motor connected to the target ball screw is collected, and the real-time torque order sequence is determined based on the real-time motor torque.

[0040] The real-time screw engagement frequency, real-time nut engagement frequency, and real-time rolling element engagement frequency are determined based on the real-time screw rotation speed, real-time screw nominal diameter, real-time ball nominal diameter, real-time screw thread helix angle, and real-time contact angle of the target ball screw.

[0041] The real-time torque order sequence, the real-time ball screw engagement frequency, the real-time nut engagement frequency, and the real-time rolling element engagement frequency are input into a pre-trained fault detection model to determine the fault information of the target ball screw.

[0042] The aforementioned ball screw fault detection method, device, and computer equipment acquire the real-time motor torque of the servo motor connected to the target ball screw, and determine the real-time torque order sequence based on the real-time motor torque. Based on the real-time screw rotation speed, real-time nominal screw diameter, real-time nominal ball diameter, real-time screw thread helix angle, and real-time contact angle, determine the real-time screw engagement frequency, real-time nut engagement frequency, and real-time rolling element engagement frequency. The real-time torque order sequence, real-time screw engagement frequency, real-time nut engagement frequency, and real-time rolling element engagement frequency are input into a pre-trained fault detection model to determine the fault information of the target ball screw. This solves the problems of high fault detection costs and the inability to obtain timely and accurate fault information for ball screws. The above-described solution, when detecting faults in ball screws, comprehensively considers the ball screw rotation speed, nominal screw diameter, nominal ball diameter, screw thread helix angle, contact angle, and the operating state of the connected servo motor. Based on the real-time screw rotation speed, nominal screw diameter, nominal ball diameter, screw thread helix angle, and contact angle of the target ball screw, it determines the real-time screw engagement frequency, nut engagement frequency, and rolling element engagement frequency, as well as the real-time torque order sequence of the servo motor. Through a pre-trained fault detection model, it determines the fault information of the ball screw. This eliminates the need for large-scale fault detection equipment, improving fault detection efficiency, accuracy, and cost. Furthermore, by utilizing servo motor torque and speed signals to construct a data-driven and mechanism-based ball screw pair fault diagnosis model, it eliminates the need for external sensors, reduces monitoring costs, and minimizes false alarms caused by abnormal data from external sensors. Attached Figure Description

[0043] Figure 1 This is an application environment diagram of a ball screw fault detection method in one embodiment;

[0044] Figure 2 This is a flowchart illustrating a fault detection method for a ball screw in one embodiment.

[0045] Figure 3 This is a flowchart illustrating a method for determining fault information of a target ball screw in one embodiment.

[0046] Figure 4 This is a flowchart illustrating a fault detection method for a ball screw in another embodiment;

[0047] Figure 5 This is a structural block diagram of a ball screw fault detection device in one embodiment;

[0048] Figure 6 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0049] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0050] The ball screw fault detection method provided in this application embodiment can be applied to, for example, Figure 1 In the application environment shown, terminal 102 communicates with server 104 via a network. A data storage system can store the data that server 104 needs to process. The data storage system can be integrated onto server 104 or placed on a cloud or other network server. Server 104 collects the real-time motor torque of the servo motor connected to the target ball screw, and determines the real-time torque order sequence based on the real-time motor torque; it determines the real-time screw engagement frequency, real-time nut engagement frequency, and real-time rolling element engagement frequency based on the real-time screw rotation speed, real-time screw nominal diameter, real-time ball nominal diameter, real-time screw thread helix angle, and real-time contact angle of the target ball screw; it inputs the real-time torque order sequence, the real-time screw engagement frequency, the real-time nut engagement frequency, and the real-time rolling element engagement frequency into a pre-trained fault detection model to determine the fault information of the target ball screw, and sends the fault information of the target ball screw to terminal 102 via the communication network. The terminal 102 can be, but is not limited to, various personal computers, laptops, smartphones, tablets, IoT devices, and portable wearable devices. IoT devices can include smart speakers, smart TVs, smart air conditioners, and smart in-vehicle systems. Portable wearable devices can include smartwatches, smart bracelets, and head-mounted devices. The server 104 can be implemented using a standalone server or a server cluster consisting of multiple servers.

[0051] In one embodiment, such as Figure 2 As shown, a fault detection method for ball screws is provided. This embodiment illustrates the method by applying it to a terminal. It is understood that this method can also be applied to a server, or to a system including both a terminal and a server, and implemented through interaction between the terminal and the server. In this embodiment, the method includes the following steps:

[0052] S210: Collect the real-time motor torque of the servo motor connected to the target ball screw, and determine the real-time torque order sequence based on the real-time motor torque.

[0053] A ball screw is a mechanical transmission device composed of a screw, nut, and balls. It converts rotary motion into linear motion through the rolling of the balls. The ball screw achieves high-precision linear motion control through the rolling of its internal balls, while a servo motor provides the driving power. The two work together to meet the high precision and efficiency requirements of CNC machine tools and automated equipment. The target ball screw refers to the ball screw to be fault-detected. The real-time torque order sequence refers to the signal sequence composed of a fixed number of motor torque signals collected per revolution of the servo motor. For example, the fixed number of motor torque signals collected per revolution of the servo motor can be used as the torque order signal. The final determined torque order signal collected after n revolutions of the servo motor can be used as the real-time torque order sequence, where n is an integer greater than or equal to 1.

[0054] Specifically, the real-time motor torque of the servo motor connected to the target ball screw is acquired through a PLC (Programmable Logic Controller). Based on the servo motor's rotational speed, the real-time motor torque is converted into a real-time torque order sequence with equal-angle sampling. Equal-angle torque acquisition is a measurement method that synchronously acquires torque data by fixing the number of sampling points per revolution, suitable for dynamic torque measurement of rotating machinery. Its core lies in ensuring that the number of sample points acquired per revolution is the same at different speeds, thereby eliminating signal aliasing and spectral tailing problems caused by speed fluctuations.

[0055] For example, the real-time torque order sequence can be represented by formula (1):

[0056] (1)

[0057] in, For real-time torque order sequence, This refers to the torque order signal of the servo motor acquired when the servo motor has rotated to the i-th revolution, where 1 ≤ i ≤ n; the torque order signal can be acquired at equal angles for each revolution of the servo motor. The motor torque signal of each servo motor, i.e. Contains One data point, It is an integer greater than or equal to 1; the real-time torque order sequence contains a total of n torque order signals.

[0058] S220. Determine the real-time screw engagement frequency, real-time nut engagement frequency, and real-time rolling element engagement frequency based on the real-time screw rotation speed, real-time screw nominal diameter, real-time ball nominal diameter, real-time screw thread helix angle, and real-time contact angle of the target ball screw.

[0059] The ball screw rotation speed refers to the number of rotations per minute (rpm) of the ball screw, and is an important indicator for measuring the speed of the screw's movement. The nominal diameter of the ball screw is the virtual ideal diameter at its location; the nominal diameter of the balls refers to the virtual ideal diameter at the location of the ball's center, i.e., the diameter of the line connecting the centers of all the balls, used to determine the stiffness and accuracy of the ball screw. The thread helix angle is the angle between the tangent of the helix on the screw's pitch cylinder and a plane perpendicular to the thread axis. The pitch cylinder of the ball screw refers to the diameter of the imaginary cylinder where the grooves and protrusions on the thread profile have equal width in the axial section of the thread, typically used to control the thread's fit accuracy and strength. The ball screw contact angle is usually the angle between the common normal of the contact surface between the balls and the raceway in a ball screw and the center diameter line of the balls, mainly used to describe the mechanical characteristics of ball screw drives. The ball screw meshing frequency refers to the number of meshes or the corresponding periodic frequency generated per revolution of the gear screw during rotation, determined by both the screw's rotational frequency and the number of teeth. The number of teeth on a leadscrew usually refers to the number of threads, i.e., the number of threads, indicating the density of the threads. The target rolling element meshing is explained below; the rolling element is the ball; the real-time nominal diameter of the leadscrew, the real-time nominal diameter of the balls, the real-time leadcrew thread helix angle, the real-time contact angle, and the real-time leadcrew rotational speed can be acquired through corresponding sensors. The meshing frequency is the periodic vibration frequency caused by gear meshing in a gear transmission system, determined by the product of the gear rotational frequency and the number of teeth.

[0060] For example, determining the real-time screw engagement frequency, real-time nut engagement frequency, and real-time rolling element engagement frequency based on the real-time screw rotation speed, real-time screw nominal diameter, real-time ball nominal diameter, real-time screw thread helix angle, and real-time contact angle of the target ball screw includes:

[0061] The real-time screw engagement frequency is determined based on the real-time screw rotation speed, real-time screw nominal diameter, real-time ball nominal diameter, real-time screw thread helix angle, and real-time contact angle; the real-time nut engagement frequency is determined based on the real-time screw rotation speed, real-time screw nominal diameter, real-time ball nominal diameter, real-time screw thread helix angle, and real-time contact angle; the real-time rolling element engagement frequency is determined based on the real-time screw rotation speed, real-time screw nominal diameter, real-time ball nominal diameter, and real-time contact angle.

[0062] Specifically, the formula for calculating the real-time lead screw engagement frequency is shown in formula (2):

[0063] (2)

[0064] Where n is the real-time lead screw rotation speed, d0 is the real-time nominal diameter of the lead screw, and d b For real-time nominal diameter of ball bearings, For real-time lead screw thread helix angle, For real-time contact angle, f sThis refers to the real-time lead screw engagement frequency.

[0065] The formula for calculating the real-time nut engagement frequency is shown in formula (3):

[0066] (3)

[0067] Among them, f n This represents the real-time nut engagement frequency.

[0068] The formula for calculating the real-time rolling element meshing frequency is shown in formula (4):

[0069] (4)

[0070] Among them, f b This refers to the real-time rolling element meshing frequency.

[0071] The above solution provides a real-time screw engagement frequency, a real-time nut engagement frequency, and a real-time rolling element engagement frequency, which can improve the accuracy of these frequencies.

[0072] S230. Input the real-time torque order sequence, real-time screw engagement frequency, real-time nut engagement frequency, and real-time rolling element engagement frequency into the pre-trained fault detection model to determine the fault information of the target ball screw.

[0073] The fault information for the target ball screw includes the probability of each possible fault category. These possible fault categories can include screw fault, nut fault, rolling element fault, and no fault.

[0074] The aforementioned ball screw fault detection method involves acquiring the real-time motor torque of the servo motor connected to the target ball screw, and determining the real-time torque order sequence based on the real-time motor torque. The method also determines the real-time screw engagement frequency, nut engagement frequency, and rolling element engagement frequency based on the real-time screw rotation speed, nominal screw diameter, nominal ball diameter, screw thread helix angle, and contact angle. These real-time torque order sequence, screw engagement frequency, nut engagement frequency, and rolling element engagement frequency are then input into a pre-trained fault detection model to determine the fault information of the target ball screw. This method solves the problems of high fault detection costs and the inability to obtain timely and accurate fault information for ball screws. The above-described solution, when detecting faults in ball screws, comprehensively considers the ball screw rotation speed, nominal screw diameter, nominal ball diameter, screw thread helix angle, contact angle, and the operating state of the connected servo motor. Based on the real-time screw rotation speed, nominal screw diameter, nominal ball diameter, screw thread helix angle, and contact angle of the target ball screw, it determines the real-time screw engagement frequency, nut engagement frequency, and rolling element engagement frequency, as well as the real-time torque order sequence of the servo motor. Through a pre-trained fault detection model, it determines the fault information of the ball screw. This eliminates the need for large-scale fault detection equipment, improving fault detection efficiency, accuracy, and cost. Furthermore, by utilizing servo motor torque and speed signals to construct a data-driven and mechanism-based ball screw pair fault diagnosis model, it eliminates the need for external sensors, reduces monitoring costs, and minimizes false alarms caused by abnormal data from external sensors.

[0075] In one embodiment, such as Figure 3 As shown, the real-time torque order sequence, real-time screw engagement frequency, real-time nut engagement frequency, and real-time rolling element engagement frequency are input into a pre-trained fault detection model to determine the fault information of the target ball screw, including:

[0076] S310. Input the real-time torque order sequence, real-time screw engagement frequency, real-time nut engagement frequency, and real-time rolling element engagement frequency into the pre-trained fault detection model so that the fault detection model can determine the convolution kernel operator based on the real-time torque order sequence, real-time screw engagement frequency, real-time nut engagement frequency, and real-time rolling element engagement frequency.

[0077] The convolution kernel operators include the convolution kernel operator corresponding to the real-time lead screw meshing frequency, the convolution kernel operator corresponding to the real-time nut meshing frequency, and the convolution kernel operator corresponding to the real-time rolling element meshing frequency.

[0078] For example, the calculation formula for the convolution kernel operator corresponding to the real-time lead screw engagement frequency is shown in formula (5):

[0079] (5)

[0080] in, is the convolution kernel operator corresponding to the real-time screw engagement frequency, n is the number of torque order signals in the real-time torque order sequence, and j is an imaginary number.

[0081] The calculation formula for the convolution kernel operator corresponding to the real-time nut engagement frequency is shown in formula (6):

[0082] (6)

[0083] in, This is the convolution kernel operator corresponding to the real-time nut engagement frequency.

[0084] The calculation formula for the convolution kernel operator corresponding to the real-time rolling element meshing frequency is shown in formula (7):

[0085] (7)

[0086] in, This is the convolution kernel operator corresponding to the real-time rolling body meshing frequency.

[0087] S320. Through the convolutional layer of the fault detection model, real-time fault detection indicators are determined based on the convolution kernel operator, real-time screw meshing frequency, real-time nut meshing frequency, and real-time rolling element meshing frequency.

[0088] Real-time fault detection indicators include: real-time lead screw engagement order, real-time nut engagement order, and real-time rolling element engagement order.

[0089] Specifically, through the convolutional layer of the fault detection model, the real-time screw engagement order is determined based on the convolutional kernel operator corresponding to the real-time screw engagement frequency; the real-time nut engagement order is determined based on the convolutional kernel operator corresponding to the real-time nut engagement frequency; and the real-time rolling element engagement order is determined based on the convolutional kernel operator corresponding to the real-time rolling element engagement frequency.

[0090] For example, the expression for the real-time lead screw engagement order is shown in formula (8):

[0091] (8)

[0092] in, This refers to the real-time lead screw engagement order.

[0093] The expression for the real-time nut engagement order is shown in formula (9):

[0094] (9)

[0095] This refers to the real-time nut engagement sequence.

[0096] The expression for the real-time rolling element meshing order is shown in formula (10):

[0097] (10)

[0098] in, This refers to the real-time rolling element meshing order.

[0099] S330. Through the attention network layer of the fault detection model, the target fault perception features are determined based on real-time fault detection indicators.

[0100] For example, the method for determining the target fault perception features may be as follows: determine the target input matrix based on the real-time fault detection indicators; determine the query vector matrix, key vector matrix, and value vector matrix based on the target input matrix through the attention network layer of the fault detection model; determine the attention matrix based on the query vector matrix and key vector matrix; and determine the intermediate representation vector based on the attention matrix and value vector matrix; and perform weighted processing on the real-time fault detection indicators based on the intermediate representation vector to determine the real-time fault perception features.

[0101] In this context, the query vector matrix is ​​the same as the Query matrix in the attention mechanism, the key vector matrix is ​​the same as the Key matrix in the attention mechanism, and the value vector matrix is ​​the same as the Value matrix in the attention mechanism.

[0102] Specifically, the expression for the target input matrix is ​​shown in formula (11):

[0103] (11)

[0104] Where F is the target input matrix.

[0105] The target input matrix is ​​processed by three sets of independent parameter matrices in the attention network layer of the fault detection model to generate a query vector matrix, a key vector matrix, and a value vector matrix. The attention matrix is ​​determined as shown in formula (12):

[0106] (12)

[0107] in, For attention matrix, To query the vector matrix, is the transpose of the key vector matrix, and d is the feature dimension corresponding to the attention network layer, which can be set according to actual needs.

[0108] The intermediate representation vector is determined as shown in formula (13):

[0109] (13)

[0110] in, Let V be the intermediate representation vector, and let V be the value vector matrix.

[0111] The real-time fault detection indicators are weighted based on the intermediate representation vector to determine the fault perception features.

[0112] For example, the fault perception characteristics are determined as shown in formula (14):

[0113] (14)

[0114] Where M represents the fault perception feature.

[0115] The above scheme provides a method for determining fault perception features based on an attention mechanism, which can fully explore the correlation between torque features and ball screw fault modes, thereby improving the accuracy of fault information of the target ball screw subsequently determined.

[0116] S340. Through the multi-layer sensor of the fault detection model, determine the fault information of the target ball screw based on the target fault perception characteristics.

[0117] For example, the multilayer perceptron of the fault detection model processes the target fault perception features based on the normalized activation function to determine the probability corresponding to the fault category of the target ball screw; the fault categories include screw fault, nut fault, rolling element fault, and no fault; the fault information of the target ball screw is determined according to the probability corresponding to the fault category.

[0118] The normalized activation function, also known as the softmax activation function, is mainly used in the output layer of multi-class tasks. It transforms the model's output into a probability distribution, meaning that the output value corresponding to each class is compressed to between 0 and 1, and the sum of all output values ​​is 1.

[0119] Specifically, the multilayer perceptron of the fault detection model processes the target fault perception features based on the normalized activation function to determine the probability corresponding to the fault category of the target ball screw. If there is a fault category whose probability is greater than a preset probability threshold, then the target ball screw is determined to have a fault corresponding to that fault category.

[0120] For example, if the preset probability threshold is 0.4, the probabilities corresponding to the fault categories of the target ball screw are as follows: the probability of screw fault is 0.1, the probability of nut fault is 0.6, the probability of rolling element fault is 0.2, and the probability of no fault is 0.1. Then the fault information of the target ball screw is that the target ball screw has a nut fault.

[0121] The above scheme processes the target fault perception features based on the normalized activation function to determine the probability corresponding to the fault category of the target ball screw. It can determine the probability of all possible faults of the target ball screw, and locate the fault of the target ball screw according to the probability corresponding to the fault category of the target ball screw, which can improve the fault detection accuracy and fault location efficiency of the target ball screw.

[0122] The above method, when determining the fault information of the target ball screw through the fault detection model, extracts the fault order amplitude through the convolution kernel operator, which can avoid the influence of motor speed fluctuations on the fault detection results of the target ball screw and improve the stability of the model.

[0123] In one embodiment, such as Figure 4 As shown, the above-mentioned fault detection method for ball screws also includes:

[0124] S410. Determine the sample motor torque of the servo motor connected to the sample ball screw, determine the sample torque order sequence based on the sample motor torque, and determine the sample state information of the sample ball screw.

[0125] The sample status information includes lead screw fault status, nut fault status, rolling element fault status, and normal status.

[0126] The sample ball screws include ball screws in the following states: screw failure state, nut failure state, rolling element failure state, and normal state.

[0127] Specifically, the sample motor torque of the servo motor connected to the sample ball screw is collected by the PLC. Based on the rotational speed of the servo motor, the sample motor torque is converted into a sample torque order sequence with equal angle sampling, and the sample state information of the sample ball screw is determined.

[0128] S420. The neural network model is trained based on the sample torque order sequence, sample lead screw meshing frequency, sample nut meshing frequency, sample rolling element meshing frequency, and sample state information to determine the fault detection model.

[0129] For example, model training data and model test data can be determined based on the sample torque order sequence, sample lead screw meshing frequency, sample nut meshing frequency, sample rolling element meshing frequency, and sample state information. The model training data is then used to train the neural network model to determine candidate detection models. The candidate detection models are then tested using the model test data to determine their cross-entropy loss function. If the cross-entropy loss function of the candidate detection model is less than a preset loss function threshold, then the candidate detection model is determined to be a fault detection model.

[0130] The above scheme provides a model training method for a fault detection model. Using the fault detection model for ball screw fault detection can improve the efficiency of ball screw fault detection.

[0131] In one embodiment, a neural network model is trained based on sample torque order sequence, sample lead screw meshing frequency, sample nut meshing frequency, sample rolling element meshing frequency, and sample state information to determine a fault detection model, including:

[0132] The neural network model is trained based on the sample torque order sequence, sample lead screw meshing frequency, sample nut meshing frequency, sample rolling element meshing frequency, and sample state information. During the training process, the cross-entropy loss function of the neural network model is determined in real time. The gradient descent method is used to update the model parameters of the neural network model based on the cross-entropy loss function until the neural network model converges. The trained neural network model is then used as the fault detection model.

[0133] Gradient descent is a first-order optimization algorithm that iteratively adjusts parameters to minimize an objective function. It calculates the gradient (first derivative) of the objective function and gradually adjusts the parameters in the opposite direction of the gradient, causing the function value to converge towards a local minimum. Gradient descent mainly includes three steps: parameter initialization, gradient calculation, and iterative update. Parameter initialization involves randomly selecting a starting point; gradient calculation involves determining the steepest descent direction at the current point; and iterative update involves adjusting the parameters in the negative gradient direction with a step size until the neural network model converges.

[0134] The above scheme uses gradient descent and updates the model parameters of the neural network model based on the cross-entropy loss function until the neural network model converges, which can improve the model training efficiency.

[0135] For example, based on the above embodiments, the fault detection method for ball screws includes:

[0136] The sample motor torque of the servo motor connected to the sample ball screw is acquired by a PLC. Based on the servo motor speed, the sample motor torque is converted into a sample torque order sequence with equal angle sampling, and the sample state information of the sample ball screw is determined. A neural network model is trained based on the sample torque order sequence and sample state information, and the cross-entropy loss function of the neural network model is determined in real time during training. Gradient descent is used to update the model parameters of the neural network model based on the cross-entropy loss function until the neural network model converges. The trained neural network model is then used as a fault detection model.

[0137] The real-time motor torque of the servo motor connected to the target ball screw is acquired by a PLC (Programmable Logic Controller). Based on the rotational speed of the servo motor, the real-time motor torque is converted into a real-time torque order sequence with equal angle sampling.

[0138] The real-time screw engagement frequency is determined based on the real-time screw rotation speed, real-time screw nominal diameter, real-time ball nominal diameter, real-time screw thread helix angle, and real-time contact angle; the real-time nut engagement frequency is determined based on the real-time screw rotation speed, real-time screw nominal diameter, real-time ball nominal diameter, real-time screw thread helix angle, and real-time contact angle; the real-time rolling element engagement frequency is determined based on the real-time screw rotation speed, real-time screw nominal diameter, real-time ball nominal diameter, and real-time contact angle.

[0139] The real-time torque order sequence, real-time screw engagement frequency, real-time nut engagement frequency, and real-time rolling element engagement frequency are input into a pre-trained fault detection model. This allows the fault detection model to determine the convolution kernel operator based on these parameters. Through the convolutional layers of the fault detection model, the real-time screw engagement order is determined based on the convolution kernel operator corresponding to the real-time screw engagement frequency; the real-time nut engagement order is determined based on the convolution kernel operator corresponding to the real-time nut engagement frequency; and the real-time rolling element engagement order is determined based on the convolution kernel operator corresponding to the real-time rolling element engagement frequency. The target input matrix is ​​determined based on real-time fault detection metrics. Then, through the attention network layer of the fault detection model, the query vector matrix, key vector matrix, and value vector matrix are determined based on the target input matrix. The attention matrix is ​​then determined based on the query vector matrix and key vector matrix, and an intermediate representation vector is determined based on the attention matrix and value vector matrix. The real-time fault detection metrics are weighted based on the intermediate representation vector to determine the fault perception features. Finally, the target fault perception features are processed using a multilayer perceptron of the fault detection model based on a normalized activation function to determine the probability corresponding to the fault category of the target ball screw. If the probability corresponding to a fault category is greater than a preset probability threshold, then the target ball screw is determined to have a fault corresponding to that fault category.

[0140] The aforementioned ball screw fault detection method involves acquiring the real-time motor torque of the servo motor connected to the target ball screw, and determining the real-time torque order sequence based on the real-time motor torque. The method also determines the real-time screw engagement frequency, nut engagement frequency, and rolling element engagement frequency based on the real-time screw rotation speed, nominal screw diameter, nominal ball diameter, screw thread helix angle, and contact angle. These real-time torque order sequence, screw engagement frequency, nut engagement frequency, and rolling element engagement frequency are then input into a pre-trained fault detection model to determine the fault information of the target ball screw. This method solves the problems of high fault detection costs and the inability to obtain timely and accurate fault information for ball screws. The above-described solution, when detecting faults in ball screws, comprehensively considers the ball screw rotation speed, nominal screw diameter, nominal ball diameter, screw thread helix angle, contact angle, and the operating state of the connected servo motor. Based on the real-time screw rotation speed, nominal screw diameter, nominal ball diameter, screw thread helix angle, and contact angle of the target ball screw, it determines the real-time screw engagement frequency, nut engagement frequency, and rolling element engagement frequency, as well as the real-time torque order sequence of the servo motor. Through a pre-trained fault detection model, it determines the fault information of the ball screw. This eliminates the need for large-scale fault detection equipment, improving fault detection efficiency, accuracy, and cost. Furthermore, by utilizing servo motor torque and speed signals to construct a data-driven and mechanism-based ball screw pair fault diagnosis model, it eliminates the need for external sensors, reduces monitoring costs, and minimizes false alarms caused by abnormal data from external sensors.

[0141] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0142] Based on the same inventive concept, this application also provides a ball screw fault detection device for implementing the above-mentioned ball screw fault detection method. The solution provided by this device is similar to the solution described in the above method. Therefore, the specific limitations of one or more ball screw fault detection device embodiments provided below can be found in the limitations of the ball screw fault detection method described above, and will not be repeated here.

[0143] In one embodiment, such as Figure 5 As shown, a fault detection device for ball screws is provided, comprising: a torque order sequence determination module 501, a meshing frequency determination module 502, and a fault information determination module 503, wherein:

[0144] The torque order sequence determination module 501 is used to collect the real-time motor torque of the servo motor connected to the target ball screw, and determine the real-time torque order sequence based on the real-time motor torque.

[0145] The meshing frequency determination module 502 is used to determine the real-time screw meshing frequency, real-time nut meshing frequency and real-time rolling element meshing frequency based on the real-time screw rotation speed, real-time screw nominal diameter, real-time ball nominal diameter, real-time screw thread helix angle and real-time contact angle of the target ball screw.

[0146] The fault information determination module 503 is used to input the real-time torque order sequence, the real-time screw engagement frequency, the real-time nut engagement frequency, and the real-time rolling element engagement frequency into a pre-trained fault detection model to determine the fault information of the target ball screw.

[0147] For example, the fault information determination module 503 is specifically used for:

[0148] The real-time torque order sequence, the real-time lead screw meshing frequency, the real-time nut meshing frequency, and the real-time rolling element meshing frequency are input into a pre-trained fault detection model so that the fault detection model can determine the convolution kernel operator based on the real-time torque order sequence, the real-time lead screw meshing frequency, the real-time nut meshing frequency, and the real-time rolling element meshing frequency.

[0149] The real-time fault detection index is determined by the convolutional layer of the fault detection model based on the convolutional kernel operator, the real-time lead screw meshing frequency, the real-time nut meshing frequency, and the real-time rolling element meshing frequency; the real-time fault detection index includes: real-time lead screw meshing order, real-time nut meshing order, and real-time rolling element meshing order.

[0150] The target fault perception features are determined based on the real-time fault detection index through the attention network layer of the fault detection model.

[0151] The fault information of the target ball screw is determined by the multilayer sensor of the fault detection model based on the target fault perception characteristics.

[0152] Furthermore, the fault information determination module 503 is also specifically used for:

[0153] The target input matrix is ​​determined based on the real-time fault detection index, and the query vector matrix, key vector matrix, and value vector matrix are determined based on the target input matrix through the attention network layer of the fault detection model.

[0154] The attention matrix is ​​determined based on the query vector matrix and the key vector matrix, and the intermediate representation vector is determined based on the attention matrix and the value vector matrix.

[0155] The real-time fault detection indicators are weighted based on the intermediate representation vector to determine the fault perception features.

[0156] Furthermore, the fault information determination module 503 is also specifically used for:

[0157] The multilayer perceptron of the fault detection model processes the target fault perception features based on a normalized activation function to determine the probability corresponding to the fault category of the target ball screw; the fault category includes screw fault, nut fault, rolling element fault, and no fault.

[0158] The fault information of the target ball screw is determined based on the probability corresponding to the fault category.

[0159] For example, the meshing frequency determination module 502 is specifically used for:

[0160] The real-time lead screw engagement frequency is determined based on the real-time lead screw rotation speed, real-time lead screw nominal diameter, real-time ball nominal diameter, real-time lead screw thread helix angle, and real-time contact angle.

[0161] The real-time nut engagement frequency is determined based on the real-time screw rotation speed, real-time screw nominal diameter, real-time ball nominal diameter, real-time screw thread helix angle, and real-time contact angle.

[0162] The real-time rolling element engagement frequency is determined based on the real-time lead screw rotation speed, real-time lead screw nominal diameter, real-time ball nominal diameter, and real-time contact angle.

[0163] For example, the above-mentioned ball screw fault detection device further includes:

[0164] The sample data determination module is used to determine the sample motor torque of the servo motor connected to the sample ball screw, determine the sample torque order sequence based on the sample motor torque, and determine the sample state information of the sample ball screw; the sample state information includes screw fault state, nut fault state, rolling element fault state, and normal state.

[0165] The model training module is used to train the neural network model based on the sample torque order sequence, sample lead screw meshing frequency, sample nut meshing frequency, sample rolling element meshing frequency, and the sample state information to determine the fault detection model.

[0166] For example, the above model training module is specifically used for:

[0167] The neural network model is trained based on the sample torque order sequence, sample lead screw meshing frequency, sample nut meshing frequency, sample rolling element meshing frequency, and the sample state information, and the cross-entropy loss function of the neural network model is determined in real time during the training process.

[0168] The gradient descent method is used to update the model parameters of the neural network model based on the cross-entropy loss function until the neural network model converges. The trained neural network model is then used as the fault detection model.

[0169] Each module in the aforementioned ball screw fault detection device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in the processor of a computer device in hardware form or independent of it, or stored in the memory of a computer device in software form, so that the processor can call and execute the operations corresponding to each module.

[0170] In one embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be as follows: Figure 6As shown, the computer device includes a processor, memory, input / output interface, communication interface, display unit, and input device. The processor, memory, and input / output interface are connected via a system bus, and the communication interface, display unit, and input device are also connected to the system bus via the input / output interface. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The input / output interface is used for exchanging information between the processor and external devices. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, mobile cellular networks, NFC (Near Field Communication), or other technologies. When executed by the processor, the computer program implements a method for detecting faults in a ball screw. The display unit is used to form a visually visible image and can be a display screen, projection device, or virtual reality imaging device. The display screen can be an LCD screen or an e-ink screen. The input device of the computer device can be a touch layer covering the display screen, or buttons, trackballs, or touchpads set on the casing of the computer device, or external keyboards, touchpads, or mice, etc.

[0171] Those skilled in the art will understand that Figure 6 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0172] In one embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to perform the following steps:

[0173] Step 1: Collect the real-time motor torque of the servo motor connected to the target ball screw, and determine the real-time torque order sequence based on the real-time motor torque.

[0174] Step 2: Determine the real-time screw engagement frequency, real-time nut engagement frequency, and real-time rolling element engagement frequency based on the target ball screw's real-time screw rotation speed, real-time screw nominal diameter, real-time ball nominal diameter, real-time screw thread helix angle, and real-time contact angle.

[0175] Step 3: Input the real-time torque order sequence, real-time screw engagement frequency, real-time nut engagement frequency, and real-time rolling element engagement frequency into the pre-trained fault detection model to determine the fault information of the target ball screw.

[0176] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, the computer program performing the following steps when executed by a processor:

[0177] Step 1: Collect the real-time motor torque of the servo motor connected to the target ball screw, and determine the real-time torque order sequence based on the real-time motor torque.

[0178] Step 2: Determine the real-time screw engagement frequency, real-time nut engagement frequency, and real-time rolling element engagement frequency based on the target ball screw's real-time screw rotation speed, real-time screw nominal diameter, real-time ball nominal diameter, real-time screw thread helix angle, and real-time contact angle.

[0179] Step 3: Input the real-time torque order sequence, real-time screw engagement frequency, real-time nut engagement frequency, and real-time rolling element engagement frequency into the pre-trained fault detection model to determine the fault information of the target ball screw.

[0180] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, performs the following steps:

[0181] Step 1: Collect the real-time motor torque of the servo motor connected to the target ball screw, and determine the real-time torque order sequence based on the real-time motor torque.

[0182] Step 2: Determine the real-time screw engagement frequency, real-time nut engagement frequency, and real-time rolling element engagement frequency based on the target ball screw's real-time screw rotation speed, real-time screw nominal diameter, real-time ball nominal diameter, real-time screw thread helix angle, and real-time contact angle.

[0183] Step 3: Input the real-time torque order sequence, real-time screw engagement frequency, real-time nut engagement frequency, and real-time rolling element engagement frequency into the pre-trained fault detection model to determine the fault information of the target ball screw.

[0184] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of related data must comply with the relevant laws, regulations and standards of the relevant countries and regions.

[0185] Those skilled in the art will understand that all or part of the processes in 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 described above. 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.

[0186] 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.

[0187] 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 patent 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 method for fault detection of a ball screw, characterized in that, include: The real-time motor torque of the servo motor connected to the target ball screw is collected, and the real-time torque order sequence is determined based on the real-time motor torque. Based on the real-time screw rotation speed, real-time screw nominal diameter, real-time ball nominal diameter, real-time screw thread helix angle, and real-time contact angle of the target ball screw, determine the real-time screw engagement frequency, real-time nut engagement frequency, and real-time rolling element engagement frequency. The real-time torque order sequence, the real-time ball screw engagement frequency, the real-time nut engagement frequency, and the real-time rolling element engagement frequency are input into a pre-trained fault detection model to determine the fault information of the target ball screw.

2. The method according to claim 1, characterized in that, The real-time torque order sequence, the real-time ball screw engagement frequency, the real-time nut engagement frequency, and the real-time rolling element engagement frequency are input into a pre-trained fault detection model to determine the fault information of the target ball screw, including: The real-time torque order sequence, the real-time lead screw meshing frequency, the real-time nut meshing frequency, and the real-time rolling element meshing frequency are input into a pre-trained fault detection model so that the fault detection model can determine the convolution kernel operator based on the real-time torque order sequence, the real-time lead screw meshing frequency, the real-time nut meshing frequency, and the real-time rolling element meshing frequency. The real-time fault detection index is determined by the convolutional layer of the fault detection model based on the convolutional kernel operator, the real-time screw engagement frequency, the real-time nut engagement frequency, and the real-time rolling element engagement frequency; the real-time fault detection index includes: real-time screw engagement order, real-time nut engagement order, and real-time rolling element engagement order. The target fault perception features are determined by using the attention network layer of the fault detection model and the real-time fault detection indicators. The fault information of the target ball screw is determined by the multilayer sensor of the fault detection model based on the target fault perception characteristics.

3. The method according to claim 2, characterized in that, Based on the real-time fault detection metrics, the attention network layer of the fault detection model determines the target fault perception features, including: The target input matrix is ​​determined based on the real-time fault detection index, and the query vector matrix, key vector matrix, and value vector matrix are determined based on the target input matrix through the attention network layer of the fault detection model. The attention matrix is ​​determined based on the query vector matrix and the key vector matrix, and the intermediate representation vector is determined based on the attention matrix and the value vector matrix. Based on the intermediate representation vector, the real-time fault detection indicators are weighted to determine the fault perception features.

4. The method according to claim 2, characterized in that, The fault information of the target ball screw is determined based on the target fault detection features using the multilayer perceptron of the fault detection model, including: The multilayer perceptron of the fault detection model processes the target fault perception features based on a normalized activation function to determine the probability corresponding to the fault category of the target ball screw; the fault category includes screw fault, nut fault, rolling element fault, and no fault. The fault information of the target ball screw is determined based on the probability corresponding to the fault category.

5. The method according to claim 1, characterized in that, Based on the real-time screw rotation speed, real-time screw nominal diameter, real-time ball nominal diameter, real-time screw thread helix angle, and real-time contact angle of the target ball screw, determine the real-time screw engagement frequency, real-time nut engagement frequency, and real-time rolling element engagement frequency, including: The real-time lead screw engagement frequency is determined based on the real-time lead screw rotation speed, real-time lead screw nominal diameter, real-time ball nominal diameter, real-time lead screw thread helix angle, and real-time contact angle. The real-time nut engagement frequency is determined based on the real-time screw rotation speed, real-time screw nominal diameter, real-time ball nominal diameter, real-time screw thread helix angle, and real-time contact angle. The real-time rolling element meshing frequency is determined based on the real-time lead screw rotation speed, real-time lead screw nominal diameter, real-time ball nominal diameter, and real-time contact angle.

6. The method according to claim 1, characterized in that, Also includes: The sample motor torque of the servo motor connected to the sample ball screw is determined, the sample torque order sequence is determined based on the sample motor torque, and the sample state information of the sample ball screw is determined; the sample state information includes screw fault state, nut fault state, rolling element fault state, and normal state. The neural network model is trained based on the sample torque order sequence, sample lead screw meshing frequency, sample nut meshing frequency, sample rolling element meshing frequency, and the sample state information to determine the fault detection model.

7. The method according to claim 6, characterized in that, The neural network model is trained based on the sample torque order sequence, sample lead screw meshing frequency, sample nut meshing frequency, sample rolling element meshing frequency, and the sample state information to determine the fault detection model, including: The neural network model is trained based on the sample torque order sequence, sample lead screw meshing frequency, sample nut meshing frequency, sample rolling element meshing frequency, and the sample state information, and the cross-entropy loss function of the neural network model is determined in real time during the training process. The gradient descent method is used to update the model parameters of the neural network model based on the cross-entropy loss function until the neural network model converges. The trained neural network model is then used as the fault detection model.

8. A fault detection device for ball screws, characterized in that, The fault detection device for the ball screw includes: The torque order sequence determination module is used to collect the real-time motor torque of the servo motor connected to the target ball screw, and determine the real-time torque order sequence based on the real-time motor torque. The meshing frequency determination module is used to determine the real-time screw meshing frequency, real-time nut meshing frequency, and real-time rolling element meshing frequency based on the real-time screw rotation speed, real-time screw nominal diameter, real-time ball nominal diameter, real-time screw thread helix angle, and real-time contact angle of the target ball screw. The fault information determination module is used to input the real-time torque order sequence, the real-time ball screw engagement frequency, the real-time nut engagement frequency, and the real-time rolling element engagement frequency into a pre-trained fault detection model to determine the fault information of the target ball screw.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 7.

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

Images