Machining Fault Detection Methods and Equipment
By using a sliding window to perform multiple samplings and feature value comparisons under the target working conditions of the machine tool, the problem of insufficient data in machining fault detection is solved, and a more efficient fault detection effect is achieved.
Patent Information
- Application Number
- CN202311804236.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2023-12-25
- Publication Date
- 2026-11-10
- Estimated Expiration
- 2043-12-25
AI Technical Summary
Existing technologies suffer from poor fault detection results due to insufficient effective data duration during machining processes, making it difficult to achieve high-resolution fault detection.
By acquiring the target signal of the machine tool under the target working condition, multiple samples are taken using a preset sliding window to determine multiple detection signals, and the feature value of each detection signal is compared with the preset threshold boundary to determine whether there is a machine tool processing fault.
By maximizing the use of data within a limited data timeframe, the impact of insufficient effective data on the detection results is effectively avoided, thereby improving the accuracy and reliability of machining fault detection.
Smart Images

Figure CN120244701B_ABST
Abstract
Description
Technical Field
[0001] This application relates to the technical field of machining, and in particular to a machining fault detection method and equipment. Background Technology
[0002] In machining processes such as turning, milling, and boring, various malfunctions are unavoidable, such as tool breakage and chipping, which affect machining efficiency and quality. Therefore, it is usually necessary to monitor the machine tool's status in real time during machining to provide timely warnings when machine tool malfunctions occur and avoid unnecessary losses.
[0003] Currently, machining fault detection solutions mainly rely on eigenvalue boundaries for control. However, due to the numerous factors affecting eigenvalues during machining, fluctuations in eigenvalues are difficult to interpret. Therefore, high-resolution data is required to ensure the reliability of the detection results. Consequently, the solution is not effective when detecting faults with small amounts of data. Furthermore, in most cases, the effective duration of relevant data during the machining process is often insufficient, thus affecting the effectiveness of machining fault detection. Summary of the Invention
[0004] One objective of this application is to provide a machining fault detection method to solve the technical problem that the detection effect is affected by insufficient effective duration of relevant data when performing machining fault detection in related technologies.
[0005] In a first aspect, embodiments of this application provide a machining fault detection method, comprising: acquiring a target signal of a machine tool under a target operating condition; sampling the target signal multiple times through a preset sliding window to determine multiple detection signals, wherein the total length of the multiple detection signals is greater than the length of the target signal; comparing the feature value of each detection signal with a preset threshold boundary to determine whether a machine tool machining fault exists, wherein the preset threshold boundary is determined based on a fault-free signal of the machine tool under the target operating condition.
[0006] In conjunction with the first aspect, in one possible implementation, acquiring the target signal of the machine tool under the target working condition includes: acquiring the machining signal of the machine tool; determining the start time and end time of the machining under the target working condition; determining an effective machining time between the start time and end time according to a preset time interval; identifying the machine tool machining signal within the effective machining time and determining a target time period in which the signal amplitude is within a preset amplitude range; and extracting the machining signal within the target time period as the target signal.
[0007] In conjunction with the first aspect, in one possible implementation, the step of performing data sampling operations on the target signal through a preset sliding window to determine multiple detection signals includes: determining the sampling length and sampling interval of the sliding window; and sequentially extracting the detection signals from the target signal according to the sampling length and sampling interval of the sliding window, wherein the length of each detection signal is the sampling length of the sliding window.
[0008] In conjunction with the first aspect, in one possible implementation, the step of sequentially extracting the detection signal from the target signal according to the sampling length and sampling interval includes: sequentially determining a plurality of start times and a plurality of end times, wherein each start time corresponds to an end time, the interval between adjacent start times is the sampling interval, and the interval between each start time and its corresponding end time is the sampling length; and sequentially extracting the detection signal from the target signal according to each start time and its corresponding end time.
[0009] In conjunction with the first aspect, in one possible implementation, comparing the feature value of each of the detected signals with a preset threshold boundary to determine whether a machine tool processing fault exists includes: each time a detection signal is determined, inputting the detection signal into a preset first model, the first model being used to extract the feature value of the detection signal and compare it with the preset threshold boundary; determining, based on the output of the first model, whether the feature value is less than the preset threshold boundary; if the feature value is less than the preset threshold boundary, determining that a machine tool processing fault exists; if the feature value is greater than the preset threshold boundary, determining that no machine tool processing fault exists.
[0010] In conjunction with the first aspect, in one possible implementation, before inputting the detection signal to the first model each time a detection signal is determined, the method further includes: acquiring historical processing signals of the machine tool; creating a dataset based on the historical processing signals; and training a preset initial model using the dataset to obtain the first model.
[0011] In conjunction with the first aspect, in one possible implementation, the dataset includes a training set and a test set. The step of training a preset initial model based on the dataset to obtain the first model includes: training the preset initial model based on the training set to obtain a target model; evaluating the target model using the test set and a preset loss function to determine the accuracy of the target model; and if the accuracy is lower than a preset accuracy threshold, updating the hyperparameters of the target model to perform multiple iterative training iterations until the accuracy of the target model is higher than the accuracy threshold, thereby obtaining the first model.
[0012] In conjunction with the first aspect, in one possible implementation, before comparing the feature value of each of the detected signals with a preset threshold boundary to determine whether a machine tool processing fault exists, the method further includes: acquiring a fault-free signal of the machine tool under the target operating condition, the fault-free signal including multiple reference signal segments; performing a data sampling operation on each of the reference signal segments according to a preset sampling rate to obtain calibration signal segments, the number of calibration signal segments being equal to the number of detected signals; and calculating the feature value of each calibration signal segment to determine the preset threshold boundary.
[0013] In conjunction with the first aspect, in one possible implementation, the step of calculating the feature value of each calibration signal segment to determine the preset threshold boundary includes: determining the waveform of each calibration signal segment; extracting waveforms from the waveforms of each calibration signal segment according to a preset waveform sampling period and waveform sampling length to determine a subset of waveforms of the calibration signal segment; calculating the feature value of each calibration signal segment based on the subset of waveforms of each calibration signal segment; and determining the preset threshold boundary based on the feature value.
[0014] In a second aspect, embodiments of this application also propose a machining fault detection device, comprising: The data acquisition module is used to acquire the target signal of the machine tool under the target working condition; The data sampling module is used to sample the target signal multiple times through a preset sliding window to determine multiple detection signals, wherein the total length of the multiple detection signals is greater than the length of the target signal; The data comparison module is used to compare the feature value of each of the detected signals with a preset threshold boundary to determine whether there is a machine tool processing fault. The preset threshold boundary is determined based on the fault-free signal of the machine tool under the target working condition.
[0015] In a third aspect, embodiments of this application also provide a computer device including a memory and a processor, the memory being connected to the processor, the processor being configured to execute one or more computer programs stored in the memory, the processor causing the computer device to perform the method as described in the first aspect when executing the one or more computer programs.
[0016] In a fourth aspect, embodiments of this application also provide a computer-readable storage medium storing a computer program, the computer program including program instructions that, when executed by a processor, cause the processor to perform the method described in the first aspect.
[0017] The embodiments of this application can achieve the following technical effects: The method proposed in this application, when used for machining fault detection, after acquiring the target signal of the machine tool under the target operating condition, performs multiple samplings through a preset sliding window to determine multiple detection signals with a total length greater than the original target signal length. By comparing the feature value of each detection signal with a preset threshold boundary, the existence of a machining fault is finally determined. Compared with related technologies, the method proposed in this application, after determining the target signal to be detected, performs multiple samplings through a sliding window, thereby extracting multiple detection signals for feature value comparison within a limited data duration, maximizing the utilization of data duration, and effectively avoiding the impact of insufficient effective data on the detection results in related technologies. Attached Figure Description
[0018] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the description of the embodiments of this application will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0019] Figure 1 This is a schematic diagram of the structure of a machining fault detection system provided in an embodiment of this application; Figure 2 A schematic flowchart illustrating a machining fault detection method provided in an embodiment of this application; Figure 3 This is a schematic diagram of the structure of a machining fault detection device provided in an embodiment of this application; Figure 4 This is a schematic diagram of the structure of a computer device provided in an embodiment of this application. Detailed Implementation
[0020] 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. All other embodiments obtained by those skilled in the art based on the embodiments in this application without inventive effort are within the scope of protection of this application.
[0021] It should be noted that, unless there is a conflict, the various features in the embodiments of this application can be combined with each other, all of which are within the protection scope of this application. Furthermore, although functional modules are divided in the device schematic diagram and a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in a different order than the module division in the device or the order in the flowchart. Moreover, the terms "first," "second," and "third" used in this application do not limit the data or execution order, but only distinguish identical or similar items with essentially the same function and effect.
[0022] To better explain this application, a machining fault detection system is first proposed.
[0023] like Figure 1 As shown, the machining fault detection system 100 includes multiple CNC machine tools 11 and multiple host computers 12. Among them, the CNC machine tools 11 are used to process workpieces, such as performing milling, drilling, reaming, boring, tapping or turning on various workpieces.
[0024] In some embodiments, the CNC machine tool 11 includes any suitable type of machine tool, such as a vertical CNC machine tool or a horizontal CNC machine tool.
[0025] The host computer 12 is used to collect various machining signals from the CNC machine tool 11 during workpiece machining. For example, the host computer 12 can collect machining signals from the CNC machine tool 11 during workpiece machining (such as turning, milling, and boring), including the physical state, mechanical effects, cutting parameters, lubrication, and cooling conditions of the CNC machine tool 11 during operation. Furthermore, the host computer 12 can also collect specific signals generated by the CNC machine tool 11 during workpiece machining under specific working conditions, such as machining signals from the CNC machine tool 11 under specific conditions and environments.
[0026] It is understood that those skilled in the art can configure the hardware architecture and software design of the host computer 12 according to the processing signals required by specific business needs. There is no limitation on the specific composition of the host computer 12. For example, if the process data includes power signals, the host computer 12 should be configured with current sensors and voltage sensors; if the process data includes triaxial acceleration signals, the host computer 12 should be configured with triaxial acceleration sensors, which can be installed in the processing area.
[0027] It is understood that the CNC machine tool 11 or host computer 12 involved in this embodiment can be configured with any suitable architecture.
[0028] Based on the above Figure 1 The machining fault detection system shown presents a machining fault detection method that can be applied to... Figure 1 The host computer is shown. For details, please refer to... Figure 2 , Figure 2 The diagram shown is a flowchart of a machining fault detection method, including: Step S10: Obtain the target signal of the machine tool under the target working condition; As those skilled in the art will understand, operating conditions refer to the collection of various conditions and environments under which a machine tool operates during actual machining. These conditions indicate the physical state, mechanical effects, cutting parameters, lubrication, and cooling status of the machine tool during operation. Specifically, the operating state of a machine tool is affected by factors such as cutting force, cutting speed, and feed rate. These factors further cause the various cutting tools mounted on the machine tool to operate according to a predetermined strategy, generating vibrations, deformations, and impact forces, ultimately determining the machining results of the machine tool.
[0029] In this embodiment, the target working condition refers to a specific working condition predetermined during the machine tool machining process. This specific working condition is used to constrain the machine tool to operate under specific conditions and environments to complete the expected workpiece machining task. Correspondingly, the target signal refers to the machining signal when the machine tool is in its target working condition, used to indicate the actual working state of the lathe.
[0030] In some embodiments, after determining the target working condition, sensors are installed on the machine tool to collect target signals under the target working condition. The sensors installed on the machine tool transmit the collected machining signals to the host computer via wireless or wired communication according to the signal acquisition time indicated by the target working condition. The host computer stores the machining signals according to different working conditions. The sensors may include accelerometers (such as piezoelectric accelerometers) to collect triaxial acceleration during machine tool machining, and may also include power sensors to collect power parameters (such as voltage, current, or actual power of the machine tool during operation).
[0031] In other embodiments, before determining the target operating condition, all machining signals generated by the machine tool throughout the entire machining process are collected using a combination of multiple sensors. These machining signals are transmitted to a host computer, which then classifies them according to their corresponding time periods and extracts the machining signal corresponding to the target operating condition as the target signal. For example, data processing software such as MATLAB can be used to analyze the time series corresponding to the machining signals to extract the machining signals corresponding to different operating conditions.
[0032] Step S20: The target signal is sampled multiple times through a preset sliding window to determine multiple detection signals, wherein the total length of the multiple detection signals is greater than the length of the target signal; It is easy to understand that a sliding window is used to extract continuous signal streams. By reasonably setting the parameters and sliding rules of the sliding window, the generated subsequence data can meet expectations and achieve the effect of specific data extraction.
[0033] In this embodiment, the preset sliding window essentially indicates a data extraction method in which the implementer sets reasonable parameters and sliding rules based on the characteristics of the target signal before implementation. Because the effective duration of the target signal is short, direct feature extraction or fault detection is ineffective and cannot accurately reflect the true working state of the machine tool. Therefore, this embodiment uses a sliding window for multiple sampling. Since the total length of the multiple detection signals is greater than the length of the target signal, this embodiment is equivalent to multiple reuses of the target signal, extending the length of each detection signal used for detection. This effectively improves the utilization rate of the target signal with limited data, thereby improving the detection effect.
[0034] Step S30: Compare the feature value of each of the detected signals with a preset threshold boundary to determine whether there is a machine tool processing fault. The preset threshold boundary is determined based on the fault-free signal of the machine tool under the target working condition.
[0035] It should be noted that the characteristic value of the detection signal refers to a numerical value or parameter used to characterize the nature and characteristics of the detection signal, representing key information when the detection signal is used for machining fault detection. This embodiment does not limit the specific types of characteristic values. Due to the different actual application scenarios and the different definitions of target working conditions by implementers, characteristic values can indicate different properties of the detection signal from multiple dimensions. However, once the target working condition is determined, the characteristic value should be able to reflect the information most relevant to the machining fault under that target working condition.
[0036] By way of example and not limitation, this embodiment provides the following commonly used eigenvalue forms: time-domain eigenvalues, including the signal's mean, variance, peak value, pulse width, etc., used to describe the signal's changes and waveform characteristics over time; frequency-domain eigenvalues, including the signal's spectral distribution, frequency components, spectral peak values, etc., used to describe the signal's frequency characteristics and frequency composition; statistical eigenvalues, including the signal's autocorrelation function, cross-correlation function, power spectral density, etc., used to describe the signal's statistical properties and correlation; energy eigenvalues, including the signal's energy distribution, energy density, energy spectrum, etc., used to describe the distribution of the signal's energy at different frequencies or time periods; peak eigenvalues, including the signal's peak amplitude, the time or frequency of peak occurrence, etc., used to describe the location and amplitude of the signal's extreme points or peak values.
[0037] For different feature value forms, different preset threshold boundaries can be set accordingly. In most cases, it is considered that a machine tool fault occurs when the feature value is less than the preset threshold boundary, and no machine tool fault occurs when the feature value is greater than the preset threshold boundary.
[0038] Furthermore, in the above embodiments, acquiring the target signal of the machine tool under the target working condition includes: acquiring the machining signal of the machine tool; determining the start time and end time of the machining under the target working condition; determining the effective machining time between the start time and end time according to a preset time interval; identifying the machine tool machining signal within the effective machining time and determining the target time period where the signal amplitude is within a preset amplitude range; and extracting the machining signal within the target time period as the target signal.
[0039] Among them, the machining signal of the machine tool refers to all machining signals of the machine tool under all possible working conditions, including the target signal set as the target working condition. The start time and end time of the target working condition refer to the time period corresponding to the target working condition. The preset time interval is a fixed duration, which is used to determine the effective machining time of the machine tool during actual machining near the time points corresponding to the start time and end time of machining. The preset amplitude range refers to a fixed waveform amplitude range.
[0040] It is easy to understand that in actual machine tool processing scenarios, the start and end times of the target processing condition are not actually equivalent to the effective processing time of the machine tool processing the workpiece. The processing signals generated during the tool's entry and exit and the time when the tool contacts the workpiece but has not started processing should not be regarded as the effective processing time of the workpiece. Therefore, the detection of machining faults should avoid the time period when the tool enters and exits.
[0041] In this embodiment, by setting a time interval, the time period unrelated to workpiece processing is excluded within the start and end processing times. For example, assuming the machine tool starts processing at time t1 and ends processing at time t2, and the preset time interval is ts, then t1+ts can be determined as the start of the effective processing time and t2-ts as the end of the effective processing time, thus obtaining the effective processing time.
[0042] In some embodiments, machine tool processing signals within the effective processing time are acquired by setting a start sampling time and an end sampling time. Assuming that the start sampling time is ta seconds after the start processing time t1 and the end sampling time is tb seconds after the start processing time t1, then ta < tb ≤ t2 - t1, to ensure that the tool is completely within the material for processing.
[0043] Further, in the above embodiments, the step of performing data sampling operations on the target signal through a preset sliding window to determine multiple detection signals includes: determining the sampling length and sampling interval of the sliding window; and sequentially extracting detection signals from the target signal according to the sampling length and sampling interval of the sliding window, wherein the length of each detection signal is the sampling length of the sliding window. The step of sequentially extracting detection signals from the target signal according to the sampling length and sampling interval includes: sequentially determining multiple start times and multiple end times, wherein each start time corresponds to an end time, the interval between adjacent start times is the sampling interval, and the interval between each start time and its corresponding end time is the sampling length; and sequentially extracting detection signals from the target signal according to each start time and its corresponding end time.
[0044] As a feasible implementation method, a queue can be defined to store the acquired target signal. When the length of the target signal is greater than a predetermined time (i.e., the sampling length mentioned above, assumed to be 1 second), the feature value is calculated and compared with a preset threshold boundary. Then, the target signal is dequeued for x milliseconds, and a new target signal is enqueued for x milliseconds. This process is repeated to form a sliding window with a sampling length of 1 second and a sampling interval of x milliseconds. Let the current time be T, the window is T-1 seconds to the left and T seconds to the right. The feature value is calculated once every x milliseconds and compared with the preset threshold boundary. If the feature value is less than the preset threshold boundary, a fault is considered to have occurred; if it is greater than the preset threshold boundary, no fault is considered to have occurred.
[0045] Further, in the above embodiments, comparing the feature value of each detection signal with a preset threshold boundary to determine whether a machine tool processing fault exists includes: each time a detection signal is determined, inputting the detection signal into a first model, the first model being used to extract the feature value of the detection signal and compare it with the preset threshold boundary; determining whether the feature value is less than the preset threshold boundary based on the output of the first model; if the feature value is less than the preset threshold boundary, then determining that a machine tool processing fault exists; if the feature value is greater than the preset threshold boundary, then determining that no machine tool processing fault exists.
[0046] The first model refers to a large model used for signal recognition and feature comparison. As an example, it can be a large model based on deep learning, including convolutional neural networks (CNN), recurrent neural networks (RNN), and support vector machines (SVM). Convolutional neural networks extract local features from the input detection signal and combine pooling layers and fully connected layers for feature extraction and classification. Recurrent neural networks are mainly used to capture the dependence between signals and time series and process time series data when identifying specific signals. Support vector machines, as a traditional machine learning model, are used to perform binary classification on the detection signal after analysis and processing, determine whether the feature value is greater than a preset threshold boundary, and output the judgment result.
[0047] In some embodiments, a preset initial model needs to be trained before using the first model, including: acquiring historical machining signals of the machine tool; creating a dataset based on the historical machining signals; and training the preset initial model using the dataset to obtain the first model.
[0048] Specifically, the steps for training a preset initial model based on the dataset include: training the preset initial model based on the training set to obtain a target model; evaluating the target model using a test set and a preset loss function to determine the accuracy of the target model; and updating the hyperparameters of the target model if the accuracy is lower than a preset accuracy threshold, and iterating the target model multiple times until the accuracy of the target model is higher than the accuracy threshold to obtain the first model.
[0049] The training set includes a large amount of labeled data, with each set containing a corresponding label or output value. The initial model learns the patterns and correlations of various signals in the data in a supervised manner, thereby making reasonable inferences given an input. Correspondingly, the sample set contains a large amount of unlabeled data used to evaluate the model's performance. In this embodiment, the loss function is used to measure the difference or error between the model's output and the actual observed values. As a feasible implementation, the Hinge function can be used to evaluate the output of the SVM classification model in the large model, which evaluates the performance of the large model by measuring the marginal difference between the output and the true label.
[0050] Furthermore, in the above embodiments, before comparing the feature value of each detection signal with the preset threshold boundary, it is also necessary to determine the preset threshold boundary, which specifically includes: acquiring the fault-free signal of the machine tool under the target working condition, the fault-free signal including multiple reference signal segments; performing data sampling operation on each reference signal segment according to a preset sampling rate to obtain a calibration signal segment, the number of calibration signal segments being equal to the number of detection signals; and calculating the feature value of each calibration signal segment to determine the preset threshold boundary.
[0051] The step of calculating the feature value of each calibration signal segment includes: determining the waveform of each calibration signal segment; extracting the waveform from the waveform of each calibration signal segment according to a preset waveform sampling period and waveform sampling length to determine the waveform subset of the calibration signal segment; calculating the feature value of each calibration signal segment based on the waveform subset of each calibration signal segment; and determining the preset threshold boundary based on the feature value.
[0052] It should be noted that the fault-free signal refers to the machining signal when the machine tool does not experience machining faults under the target working conditions. The reference signal segment refers to a segment of the fault-free signal. In order to maximize the use of data and ensure the accuracy of the preset threshold boundary, this embodiment samples n segments (n is the number of acquisition segments, preferably n>7) of signals based on a preset sampling rate, thereby obtaining n calibration signal segments. All n calibration signal segments are used to calculate the preset threshold boundary, thereby ensuring the reliability of the preset threshold boundary.
[0053] In summary, this application acquires the target signal of the machine tool under the target working condition and then performs multiple samplings through a preset sliding window to determine multiple detection signals with a total length greater than the original target signal length. By comparing the feature value of each detection signal with a preset threshold boundary, the existence of a machining fault is finally determined. Compared with related technologies, the method proposed in this application, after determining the target signal to be detected, performs multiple samplings through a sliding window, thereby extracting multiple detection signals for feature value comparison within a limited data duration. This maximizes the utilization of data duration and effectively avoids the impact of insufficient effective data on the detection results in related technologies.
[0054] As another aspect of the embodiments of this application, this application provides a machining fault detection device. The machining fault detection device can be a software module, which includes several instructions stored in a memory. A processor can access the memory, call the instructions, and execute them to complete the machining fault detection methods described in the various embodiments above.
[0055] In some embodiments, the machining fault detection device can also be constructed from hardware components. For example, the machining fault detection device can be constructed from one or more chips, which can work in coordination to complete the machining fault detection methods described in the various embodiments above. As another example, the machining fault detection device can also be constructed from various logic devices, such as general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), microcontrollers, ARM (Acorn RISC Machine) or other programmable logic devices, discrete gate or transistor logic, discrete hardware components, or any combination of these components.
[0056] Specifically, please refer to Figure 3 , Figure 3 The figure shows a schematic diagram of a machining fault detection device. As shown, the device includes: Data acquisition module 310 is used to acquire the target signal of the machine tool under the target working condition; The data sampling module 320 is used to sample the target signal multiple times through a preset sliding window to determine multiple detection signals, wherein the total length of the multiple detection signals is greater than the length of the target signal; The data comparison module 330 is used to compare the feature value of each of the detection signals with a preset threshold boundary to determine whether there is a machine tool processing fault. The preset threshold boundary is determined based on the fault-free signal of the machine tool under the target working condition.
[0057] As a feasible implementation method, the data acquisition module 310, when used to acquire the target signal of the machine tool under the target working condition, is specifically used for: acquiring the machining signal of the machine tool; determining the start time and end time of the machining under the target working condition; determining the effective machining time between the start time and end time according to a preset time interval; identifying the machine tool machining signal within the effective machining time and determining the target time period where the signal amplitude is within a preset amplitude range; and extracting the machining signal within the target time period as the target signal.
[0058] As a feasible implementation, the data sampling module 320, when performing data sampling operations on the target signal through a preset sliding window to determine multiple detection signals, is specifically used to: determine the sampling length and sampling interval of the sliding window; and sequentially extract the detection signals from the target signal according to the sampling length and sampling interval of the sliding window, wherein the length of each detection signal is the sampling length of the sliding window.
[0059] As a feasible implementation, the data sampling module 320, when extracting the detection signal sequentially from the target signal according to the sampling length and sampling interval, specifically performs the following steps: sequentially determining multiple start times and multiple end times, wherein each start time corresponds to an end time, the interval between adjacent start times is the sampling interval, and the interval between each start time and its corresponding end time is the sampling length; and sequentially extracting the detection signal from the target signal according to each start time and its corresponding end time.
[0060] As a feasible implementation, the data comparison module 320, when comparing the feature value of each detected signal with a preset threshold boundary to determine whether a machine tool processing fault exists, specifically performs the following: each time a detected signal is determined, the detected signal is input to a first model, the first model is used to extract the feature value of the detected signal and compare it with the preset threshold boundary; based on the output of the first model, it is determined whether the feature value is less than the preset threshold boundary; if the feature value is less than the preset threshold boundary, it is determined that a machine tool processing fault exists; if the feature value is greater than the preset threshold boundary, it is determined that no machine tool processing fault exists.
[0061] As a feasible implementation, the data comparison module 320, when inputting the detection signal to the first model each time a detection signal is determined, is also used to: acquire the historical processing signals of the machine tool; create a dataset based on the historical processing signals; and train a preset initial model using the dataset to obtain the first model.
[0062] As a feasible implementation method, the data comparison module 320, when used to train a preset initial model based on the dataset to obtain the first model, specifically performs the following steps: training the preset initial model based on the training set to obtain a target model; evaluating the target model using the test set and a preset loss function to determine the accuracy of the target model; and if the accuracy is lower than a preset accuracy threshold, updating the hyperparameters of the target model to perform multiple iterations of training on the target model until the accuracy of the target model is higher than the accuracy threshold, thereby obtaining the first model.
[0063] As a feasible implementation, the data comparison module 330, when comparing the feature value of each of the detected signals with a preset threshold boundary to determine whether a machine tool processing fault exists, is further configured to: acquire a fault-free signal of the machine tool under the target operating condition, the fault-free signal including multiple reference signal segments; perform a data sampling operation on each of the reference signal segments according to a preset sampling rate to obtain a calibration signal segment, the number of calibration signal segments being equal to the number of detected signals; and calculate the feature value of each calibration signal segment to determine the preset threshold boundary.
[0064] As a feasible implementation, the data comparison module 330, when calculating the feature value of each calibration signal segment to determine the preset threshold boundary, is specifically used for: determining the waveform of each calibration signal segment; extracting waveforms from the waveforms of each calibration signal segment according to a preset waveform sampling period and waveform sampling length to determine a subset of waveforms of the calibration signal segment; calculating the feature value of each calibration signal segment based on the subset of waveforms of each calibration signal segment; and determining the preset threshold boundary based on the feature value.
[0065] It should be noted that the above-described machining fault detection device can execute the machining fault detection method provided in the embodiments of this application, and has the corresponding functional modules and beneficial effects of the method. Technical details not described in detail in the embodiments of the machining fault detection device can be found in the machining fault detection method provided in the embodiments of this application.
[0066] See Figure 4 , Figure 4This is a schematic diagram of the structure of a computer device provided in an embodiment of this application. The computer device 40 includes one or more processors 41 and a memory 42. The memory 42 is connected to one or more processors 41, for example, via a bus.
[0067] Processor 41 is configured to support the computer device in performing the corresponding functions in the methods described in the above method embodiments. Processor 41 may be a central processing unit (CPU), a network processor (NP), a hardware chip, or any combination thereof. The aforementioned hardware chip may be an application-specific integrated circuit (ASIC), a programmable logic device (PLD), or a combination thereof. The aforementioned PLD may be a complex programmable logic device (CPLD), a field-programmable gate array (FPGA), a generic array logic (GAL), or any combination thereof.
[0068] Memory 42 is used to store program code, etc. Memory 42 may include volatile memory (VM), such as random access memory (RAM); memory 42 may also include non-volatile memory (NVM), such as read-only memory (ROM), flash memory, hard disk drive (HDD), or solid-state drive (SSD); memory 42 may also include combinations of the above types of memory.
[0069] The memory 42 can be used to store non-volatile software programs, non-volatile computer-executable programs, and modules, such as the program instructions / modules corresponding to the machining fault detection method in the embodiments of this application. The processor 41 executes various functional applications and data processing of the machining fault detection method and the machining fault detection device by running the non-volatile software programs, instructions, and modules stored in the memory 42, thereby realizing the functions of each module or unit of the machining fault detection method and the machining fault detection device provided in the above method embodiments.
[0070] The memory 42 may include a program storage area and a data storage area, wherein the program storage area may store the operating system and application programs required for at least one function. The data storage area may store data created based on the use of the machining fault detection device. In some embodiments, the memory 42 may optionally include memory remotely located relative to the processor 41, and this remote memory may be connected to the machining fault detection device via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.
[0071] The one or more modules are stored in the memory 42. When executed by the one or more processors 41, they perform the machining fault detection method in any of the above method embodiments. For example, they perform the method steps described in the above method embodiments to realize the functions of the modules described in the above device embodiments.
[0072] This application also provides a computer-readable storage medium storing a computer program, the computer program including program instructions, which, when executed by a computer, cause the computer to perform the method described in the foregoing embodiments.
[0073] 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 program can be stored in a computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. The storage medium can be a magnetic disk, optical disk, read-only memory (ROM), or random access memory (RAM), etc.
[0074] The above-disclosed embodiments are merely preferred embodiments of this application and should not be construed as limiting the scope of this application. Therefore, any equivalent variations made in accordance with the claims of this application shall still fall within the scope of this application.
Claims
1. A method for detecting machining faults, characterized in that, include: Acquiring a target signal from a machine tool under a target operating condition includes: acquiring the machining signal from the machine tool; determining the start and end times of the machining under the target operating condition; determining an effective machining time between the start and end times based on a preset time interval; identifying the machine tool machining signal within the effective machining time and determining a target time period where the signal amplitude is within a preset amplitude range; and extracting the machining signal within the target time period as the target signal. The target signal is sampled multiple times through a preset sliding window to determine multiple detection signals, the total length of which is greater than the length of the target signal. This includes: determining the sampling length and sampling interval of the sliding window; sequentially determining multiple start times and multiple end times, wherein each start time corresponds to an end time, the interval between adjacent start times is the sampling interval, and the interval between each start time and its corresponding end time is the sampling length; and sequentially extracting the detection signals from the target signal according to each start time and its corresponding end time. The feature value of each of the detected signals is compared with a preset threshold boundary to determine whether there is a machine tool processing fault. The preset threshold boundary is determined based on the fault-free signal of the machine tool under the target operating condition.
2. The method according to claim 1, characterized in that, The step of comparing the feature value of each detected signal with a preset threshold boundary to determine whether a machine tool processing fault exists includes: Each time a detection signal is determined, the detection signal is input to a first model, which is used to extract the feature values of the detection signal and compare them with the preset threshold boundary. Based on the output of the first model, determine whether the feature value is less than the preset threshold boundary; If the feature value is less than the preset threshold boundary, then a machine tool processing fault is determined to exist; If the feature value is greater than the preset threshold boundary, it is determined that there is no machine tool processing fault.
3. The method according to claim 2, characterized in that, Each time a detection signal is determined, the detection signal is input to the first model. Before the first model extracts the feature values of the detection signal and compares them with the preset threshold boundary, the method further includes: Acquire the historical machining signals of the machine tool; Create a dataset based on the historical processing signals; The first model is obtained by training a preset initial model using the dataset.
4. The method according to claim 3, characterized in that, The dataset includes a training set and a test set. The step of training a preset initial model based on the dataset to obtain the first model includes: The target model is obtained by training the preset initial model based on the training set. The target model is evaluated using the test set and a preset loss function to determine its accuracy. If the accuracy is lower than a preset accuracy threshold, the hyperparameters of the target model are updated to perform multiple iterations of training on the target model until the accuracy of the target model is higher than the accuracy threshold, thus obtaining the first model.
5. The method according to claim 1, characterized in that, Before comparing the feature value of each of the detected signals with a preset threshold boundary to determine whether a machine tool processing fault exists, the method further includes: Acquire a fault-free signal of the machine tool under the target operating condition, wherein the fault-free signal includes multiple reference signal segments; According to a preset sampling rate, a data sampling operation is performed on each of the reference signal segments to obtain a calibration signal segment, the number of which is equal to the number of the detected signals; Calculate the feature value of each of the calibration signal segments to determine the preset threshold boundary.
6. The method according to claim 5, characterized in that, The step of calculating the feature value of each of the calibrated signal segments to determine the preset threshold boundary includes: Determine the waveform of each of the calibration signal segments; According to the preset waveform sampling period and waveform sampling length, waveform extraction is performed on the waveform of each calibration signal segment to determine the waveform subset of the calibration signal segment; Calculate the characteristic value of each calibration signal segment based on the waveform subset of each calibration signal segment; The preset threshold boundary is determined based on the feature value.
7. A computer device, characterized in that, The device includes a memory and a processor, the memory being connected to the processor, the processor being configured to execute one or more computer programs stored in the memory, the processor causing the computer device to perform the method as described in any one of claims 1-6 when executing the one or more computer programs.
Citation Information
Patent Citations
Real-time monitoring method for waste falling state of crop shear based on sound
CN115121859A
Bearing fault detection method, device and equipment and readable storage medium
CN115307907A
