Abnormality detection method and device for tool magazine tool changing speed reducer and storage medium

Abstract

The invention discloses an anomaly detection method and device for a tool changer tool changing speed reducer and a storage medium, and relates to the technical field of tool anomaly detection.The method comprises the steps that in the machining process of a machine tool, a high-frequency operation state parameter set of the tool changing speed reducer is monitored and obtained, and a coarse anomaly risk coefficient sequence is output; the operation state parameter set of the tool changing speed reducer is obtained through fixed-point monitoring according to the second collection frequency; if the coarse abnormal risk coefficient sequence meets a first preset early warning index, outputting a fine abnormal risk coefficient sequence; and if the fine abnormal risk coefficient sequence meets a second preset early warning index, abnormal early warning is carried out on the tool changing speed reducer. The technical problems that in the prior art, due to the fact that graded early warning cannot be achieved for abnormal state detection of the tool changing speed reducer, sensitivity is insufficient, and reliability is low are solved, and the technical effects that graded early warning for the abnormal state of the tool changing speed reducer is achieved, and detection sensitivity and reliability are improved are achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of tool abnormality detection, and in particular to a method, device and storage medium for detecting abnormality of a tool magazine tool changer reducer. Background Art

[0002] In machine tool processing, tool changers are key transmission components. Abnormal operation of these components can lead to reduced machining accuracy and even equipment failure. Existing detection methods suffer from fixed monitoring frequencies and a single risk identification hierarchy, making it difficult to accurately capture equipment anomaly trends early on. A single acquisition frequency cannot balance monitoring efficiency and accuracy. Warning mechanisms with fixed thresholds are prone to false alarms or missed alerts due to noise interference. Furthermore, there is a lack of a tiered detection mechanism to enable a progressive assessment from initial risk screening to accurate confirmation.

[0003] The existing technology has the technical problem that it is unable to implement graded warning for abnormal state detection of the tool changing reducer, resulting in insufficient sensitivity and low reliability. Summary of the Invention

[0004] The present application provides a method, device and storage medium for detecting abnormalities of a tool magazine tool changer reducer, which is used to solve the technical problem in the prior art that abnormal state detection of the tool changer reducer cannot achieve graded warning, resulting in insufficient sensitivity and low reliability.

[0005] In view of the above problems, the present application provides a method, device and storage medium for detecting abnormalities of a tool magazine tool change reducer.

[0006] A first aspect of an embodiment of the present application provides a method for detecting an abnormality of a tool magazine tool change reducer, the method comprising:

[0007] During the machine tool processing process, the high-frequency operating status parameter set of the tool changing reducer is obtained by fixed-point monitoring according to the first acquisition frequency, and risk identification is performed through the abnormality detection small model to output a rough abnormality risk coefficient sequence; the operating status parameter set of the tool changing reducer is obtained by fixed-point monitoring according to the second acquisition frequency; if the rough abnormality risk coefficient sequence meets the first preset early warning index, the operating status parameter set is risk identified through the abnormality detection large model, and a fine abnormality risk coefficient sequence is output; if the fine abnormality risk coefficient sequence meets the second preset early warning index, an abnormality warning is issued for the tool changing reducer.

[0008] According to a second aspect of the embodiments of the present application, the present application provides an electronic device comprising: a processor; a memory for storing instructions executable by the processor; wherein the processor is used to execute a method for detecting an abnormality of a tool magazine tool changing reducer provided in the present application.

[0009] According to a third aspect of the embodiments of the present application, the present application provides a computer-readable storage medium storing a computer program, which is used to execute a method for detecting an abnormality of a tool magazine tool change reducer provided by the present application.

[0010] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0011] During the machining process, the system uses a first acquisition frequency to monitor the tool changer's high-frequency operating status parameters, outputting a coarse abnormality risk coefficient sequence. A second acquisition frequency is used to monitor the tool changer's operating status parameters. If the coarse abnormality risk coefficient sequence meets the first preset early warning indicator, a fine abnormality risk coefficient sequence is output. If the fine abnormality risk coefficient sequence meets the second preset early warning indicator, an abnormality warning is issued for the tool changer. This achieves a hierarchical early warning system for abnormal tool changer status, improving detection sensitivity and reliability. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0013] Figure 1 A schematic flow chart of a method for detecting an abnormality of a tool magazine tool change reducer provided in an embodiment of the present application;

[0014] Figure 2 This is a schematic diagram of the structure of an electronic device provided in this application.

[0015] Description of reference numerals: processor 21 , memory 22 , input device 23 , output device 24 . DETAILED DESCRIPTION

[0016] The present application provides a method, device and storage medium for detecting abnormalities of a tool magazine tool changer reducer, which is used to solve the technical problem in the prior art that abnormal state detection of the tool changer reducer cannot achieve graded warning, resulting in insufficient sensitivity and low reliability.

[0017] The following will be combined with the accompanying drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making any creative work are within the scope of protection of this application.

[0018] Example 1, as Figure 1 As shown, the present application provides a method for detecting abnormalities of a tool magazine tool change reducer, the method comprising:

[0019] Step S100: During the machining process, a high-frequency operating state parameter set of the tool-changing reducer is obtained by fixed-point monitoring according to a first acquisition frequency, risk identification is performed through a small abnormality detection model, and a rough abnormality risk coefficient sequence is output.

[0020] Specifically, during the machining process, the tool changer's preset operating status monitoring parameters, including several initial monitoring indicators, are first obtained. A correlation analysis is then performed to select the highest correlation indicators (less than 30%) from the initial monitoring indicators as high-frequency monitoring indicators, forming a set of high-frequency monitoring indicators. This set of high-frequency operating status parameters is then acquired through fixed-point monitoring at a relatively high first acquisition frequency. Next, using the high-frequency monitoring indicator set as a constraint, multiple sample high-frequency monitoring indicator sets are collected based on historical operation and maintenance monitoring records. The failure probability under each sample state is statistically calculated to generate a set of sample abnormality risk coefficients. This is used as training data to train a generative adversarial network until convergence, resulting in a small anomaly detection model. Because this small model only targets high-frequency monitoring indicators (a relatively small number of monitoring indicators), it exhibits high recognition efficiency but low accuracy. Finally, the sequence of high-frequency operating status parameter sets within a preset time window is input into the small anomaly detection model to complete coarse-grained risk identification. This output represents a sequence of coarse abnormality risk coefficients reflecting potential abnormality risks, providing a preliminary screening basis for subsequent high-precision identification using the large model with a complete set of indicators.

[0021] Step S200: acquiring an operating status parameter set of the tool-changing reducer through fixed-point monitoring according to a second acquisition frequency.

[0022] Specifically, based on all initial monitoring indicators (including the remaining indicators not selected as high-frequency monitoring indicators), the tool change reducer is monitored at a second acquisition frequency lower than the first acquisition frequency to obtain a set of operating status parameters containing complete indicators. At the same time, a preset time period is configured for risk trend judgment, and the sequence of crude abnormality risk coefficients within this period is extracted and analyzed for its trend, and the crude abnormality risk growth rate is calculated; the frequency compensation coefficient is generated by the formula 1 + crude abnormality risk growth rate × preset constant (≤15), and the second acquisition frequency is regularly and dynamically adjusted to automatically increase the acquisition frequency when the risk increases, enhance the monitoring sensitivity to abnormal conditions, and achieve the rational allocation of monitoring resources and dynamic adaptation to risk changes.

[0023] Step S300: If the coarse abnormal risk coefficient sequence meets the first preset warning indicator, the operating state parameter set is subjected to risk identification through the abnormality detection large model, and a fine abnormal risk coefficient sequence is output.

[0024] Specifically, if the sequence of coarse anomaly risk coefficients satisfies the first preset warning indicator including a coarse risk coefficient threshold and a coarse risk frequency threshold (where the coarse risk coefficient threshold is less than the subsequent fine risk coefficient threshold, and the coarse risk frequency threshold is less than the fine risk frequency threshold), then a large anomaly detection model is constructed based on all initial monitoring indicators. The specific process is as follows: with the initial monitoring indicators as constraints, multiple sample monitoring indicator sets and sample anomaly risk coefficient sets are collected as training data, the data is divided into K equal parts (10≤K≤30), and K training sets are obtained by selecting K times with replacement. The generative adversarial network is supervised and trained until convergence, and K large anomaly detection branches are obtained and a large anomaly detection model is constructed. Subsequently, the mean value of the crude risk coefficient overflow when the crude abnormal risk coefficient sequence meets the first preset warning indicator is calculated, and the number of branch selections P is obtained by rounding (mean value of crude risk coefficient overflow ÷ mean value of historical maximum crude risk coefficient overflow) × K. P branches are randomly selected from the K large abnormality detection branches, and risk identification is performed on the obtained operating status parameter sequence (based on the collection of all initial monitoring indicators). A more accurate precise abnormal risk coefficient sequence is output to achieve progressive detection from coarse-grained early warning to fine-grained risk identification.

[0025] Step S400: If the precise abnormality risk coefficient sequence meets a second preset warning indicator, an abnormality warning is issued to the tool-changing reducer.

[0026] Specifically, if the fine abnormality risk coefficient sequence meets the second preset warning indicator consisting of a fine risk coefficient threshold and a fine risk frequency threshold (where the fine risk coefficient threshold is higher than the coarse risk coefficient threshold in the first preset warning indicator, and the fine risk frequency threshold is higher than the coarse risk frequency threshold), it is determined that there is a real abnormality risk in the tool change reducer, and the abnormality warning mechanism is immediately triggered. This warning will alert maintenance personnel through indicator lights, pop-up windows, or text messages, accurately locating the abnormal state so that timely maintenance can be intervened to avoid machine tool shutdowns or processing quality problems caused by reducer failures, thereby achieving closed-loop management and proactive prevention and control of equipment operation risks.

[0027] In one possible implementation, step S100 further includes:

[0028] Step S110: obtaining preset operating status monitoring parameters of the tool-changing reducer, wherein the preset operating status monitoring parameters include a plurality of initial monitoring indicators.

[0029] Step S120: performing correlation analysis on the several initial monitoring indicators and the abnormal state of the tool changing reducer respectively, and selecting the initial monitoring indicators with the maximum correlation of the preset proportion as high-frequency monitoring indicators to obtain a high-frequency monitoring indicator set, wherein the preset proportion is less than 30%.

[0030] Step S130: Based on the high-frequency monitoring indicator set, a high-frequency operating status parameter set of the tool-changing reducer is acquired by fixed-point monitoring according to a first acquisition frequency.

[0031] Specifically, the preset operating status monitoring parameters for the tool-changing reducer are obtained. These parameters cover several initial monitoring indicators, such as tool change time, which normally ranges from 0.8s to 1.5s and can be used to identify faults such as jamming and poor deceleration. The motor current peak is normally between 1.5A and 3.0A, which can be used to check for abnormal increases in tool change loads. The current fluctuation range is ±0.3A, an indicator of load stability. The current duration is related to the tool change duration; abnormally long tool change times indicate increased friction resistance. The output torque is normally between 1.2 and 2.0Nm; increased torque may indicate problems such as insufficient lubrication. The root mean square (RMS) value of vibration is normally between 0.1 and 0.5g, indicating loose assembly. The spectral characteristic peak is used to detect local gear or bearing damage. The temperature rise rate is less than 2°C / minute, which can be used to determine lubrication conditions. The tool change angular displacement is accurate to ±0.5° of the specified tool position, which can determine whether there is positional deviation. These indicators together constitute the preset operating status monitoring parameters, providing basic data for subsequent analysis.

[0032] For several initial monitoring indicators of the tool change reducer obtained (such as tool change time, motor current peak, etc.), use methods such as Pearson correlation coefficient to analyze the degree of correlation between them and abnormal conditions of the tool change reducer (such as jamming, insufficient lubrication, etc.). By quantifying the correlation value, sort the correlation from high to low, and select the initial monitoring indicators with the highest correlation according to a preset ratio of less than 30%. For example, if there are 10 initial monitoring indicators, select up to 3 indicators with the highest correlation, set these indicators as high-frequency monitoring indicators, and then form a high-frequency monitoring indicator set. This operation aims to focus on key indicators that are closely related to equipment abnormalities, laying the foundation for subsequent efficient monitoring of equipment operation status and accurate identification of potential risks.

[0033] Based on the determined high-frequency monitoring indicator set (this indicator set is a set of key indicators that are most correlated with the abnormal state of the tool changer and account for less than 30% of the total, selected from the initial monitoring indicators), a relatively high first acquisition frequency (this frequency setting is intended to achieve high-frequency monitoring of key indicators to capture subtle changes in equipment operation) is used for fixed-point monitoring. During the operation of the tool changer, real-time data collection is performed on various indicators in the high-frequency monitoring indicator set (such as tool change time, motor current peak and other key indicators) at fixed time intervals or specific trigger conditions, and the collected data of various indicators are integrated to form a high-frequency operating status parameter set for the tool changer. This parameter set contains key operating status information of the equipment, and provides data support for the subsequent preliminary identification of risks through a small anomaly detection model.

[0034] In one possible implementation, step S100 further includes:

[0035] Step S140: Using the high-frequency monitoring indicator set as a constraint, based on the historical operation and maintenance monitoring records of the tool changer reducer, collect multiple sample high-frequency monitoring indicator sets, and statistically analyze the failure probability of the tool changer reducer under different sample high-frequency monitoring indicator set states, set it as the sample abnormality risk coefficient, and obtain the sample abnormality risk coefficient set.

[0036] Step S150: using the multiple sample high-frequency monitoring indicator sets and sample abnormality risk coefficient sets as training data, training the generative adversarial network until convergence, and obtaining a small anomaly detection model.

[0037] Step S160: Monitor and obtain a high-frequency operating status parameter set sequence within a preset time window, input the anomaly detection small model into a fixed point for risk identification, and output a rough anomaly risk coefficient sequence.

[0038] Specifically, with a determined set of high-frequency monitoring indicators as a constraint, multiple sample high-frequency monitoring indicator sets are collected in a targeted manner from the long-term accumulated historical operation and maintenance monitoring records of the tool changer reducer. These samples cover the operating data of the equipment under different working conditions and at different time points. For each sample high-frequency monitoring indicator set, its corresponding operating status is analyzed in detail. By studying the actual situation of the equipment failure, combined with historical maintenance records, fault reports and other information, the probability of failure of the tool changer reducer under the sample state is statistically calculated. This probability value is set as the sample abnormality risk coefficient, and the abnormal risk coefficients corresponding to multiple samples are summarized to form a sample abnormality risk coefficient set. This set is a quantitative reflection of the historical operating risk status of the equipment, and provides key basic data for the subsequent training of the generative adversarial network, so that the trained small anomaly detection model can learn the correlation pattern between the equipment operating status and the failure risk.

[0039] The obtained multiple sample high-frequency monitoring indicator sets and their corresponding sample abnormality risk coefficient sets are used as training data and input into the generative adversarial network (GAN) for supervised training. The generative adversarial network continuously optimizes the network parameters through the game process of the generator fitting the distribution of sample data and the discriminator distinguishing between real data and generated data. In the specific training process, the high-frequency monitoring indicator set is used as the input feature and the sample abnormality risk coefficient is used as the label. The network weight is adjusted through the back propagation algorithm until the simulated data distribution output by the generator is sufficiently close to the real sample distribution, and the discriminator cannot accurately distinguish between true and false, that is, the network is judged to converge. The resulting small anomaly detection model has the ability to perform coarse-grained risk identification of the tool change reducer operating status based on the high-frequency monitoring indicator set, and can quickly output a coarse abnormality risk coefficient sequence reflecting potential abnormal risks through lightweight calculations.

[0040] According to the first acquisition frequency, the high-frequency operating status parameters of the tool-changing reducer within the preset time window (such as the last 10 minutes) are continuously monitored to form a parameter set sequence containing multiple sets of high-frequency monitoring indicator data (such as collecting high-frequency monitoring indicators once per second to form a sequence of 600 sets of data). Subsequently, the parameter set sequence is input into the trained anomaly detection small model at a fixed time interval (i.e., a fixed point). Since the model is only trained based on a high-frequency monitoring indicator set (accounting for less than 30% of the initial indicators), it has lightweight computing characteristics and can quickly perform coarse-grained risk identification on the input data. The model generates a risk probability value output by the discriminator of the adversarial network, combines it with the historical threshold mapping to the corresponding coarse anomaly risk coefficient, and finally generates a coarse anomaly risk coefficient sequence reflecting the risk trend of equipment operation in chronological order, providing a preliminary basis for the subsequent judgment of whether to trigger the large model precision detection.

[0041] In one possible implementation, step S200 further includes:

[0042] Step S210: Based on the several initial monitoring indicators, obtain the operating status parameter set of the tool changing reducer through fixed-point monitoring according to a second acquisition frequency, wherein the second acquisition frequency is lower than the first acquisition frequency.

[0043] Step S220: configuring a preset time period for risk trend judgment, obtaining a sequence of crude abnormal risk coefficients within the preset time period, performing crude abnormal risk trend analysis, and outputting a crude abnormal risk growth rate.

[0044] Step S230: using 1 plus the product of the rough abnormal risk growth rate and a preset constant as a frequency compensation coefficient, wherein the preset constant is less than or equal to 15.

[0045] Step S240: regularly adjusting the second acquisition frequency according to the frequency compensation coefficient.

[0046] Specifically, based on all the initial monitoring indicators determined (covering a complete parameter system including vibration frequency, temperature, and speed), the tool change reducer is monitored at a second acquisition frequency lower than the first acquisition frequency (e.g., the first acquisition frequency is 10 times per second, and the second acquisition frequency is 1 time per minute) to obtain a set of operating status parameters containing all the initial indicators. Through differentiated frequency design, while ensuring the overall status coverage of the equipment, the monitoring resource consumption of non-critical indicators is reduced, forming a hierarchical monitoring mechanism with high frequency focusing on key indicators and low frequency covering complete parameters, laying the foundation for the subsequent dynamic adjustment of the monitoring strategy in combination with gross abnormality risks.

[0047] A preset time period (e.g., configurable time intervals such as 30 minutes or 1 hour) is pre-set to assess risk trends, and a sequence of crude anomaly risk coefficients output within that period is extracted from stored historical data. Using time series analysis methods such as sliding average, exponential smoothing, or linear regression, the changing trend of the crude anomaly risk coefficient is quantitatively analyzed, the growth rate of the risk coefficient per unit time is calculated, and the crude anomaly risk growth rate is output. This growth rate reflects the evolution of the tool changer reducer's anomaly risk over time, providing a key trend judgment basis for the subsequent dynamic adjustment of the monitoring frequency and the triggering of the fine detection mechanism.

[0048] The frequency compensation coefficient is calculated using the formula 1 + crude anomaly risk growth rate × preset constant, where the preset constant is a configurable parameter and has a value of ≤15. The design logic of this formula is: when the crude anomaly risk growth rate is positive, indicating an upward trend in equipment anomaly risk, the influence of the growth rate is amplified by the introduction of a preset constant, making the compensation coefficient greater than 1, thereby increasing the second acquisition frequency. If the growth rate is negative or zero, the compensation coefficient is ≤1, maintaining or reducing the acquisition frequency. For example, if the crude anomaly risk growth rate is 0.2 / hour and the preset constant is 10, the frequency compensation coefficient is 1 + 0.2 × 10 = 3. At this time, the second acquisition frequency will be adjusted to 3 times the original frequency, thereby achieving a dynamic response to risk changes, increasing monitoring density when risks intensify, and optimizing resource allocation efficiency.

[0049] A mechanism for regularly adjusting the second acquisition frequency based on the frequency compensation coefficient identifies abnormal trends in the crude anomaly risk coefficient sequence and uses a time series analysis algorithm to fit the crude anomaly risk growth rate. Based on this growth rate, the frequency compensation coefficient is dynamically calculated using the formula 1 + crude anomaly risk growth rate × a preset constant. This allows the full data collection frequency (i.e., the second acquisition frequency) to be adjusted based on all initial monitoring indicators. When equipment operation is stable and the risk growth rate is low, maintaining a low second acquisition frequency can save acquisition bandwidth, processing power, and storage space, reducing operation and maintenance costs. However, when a rapidly increasing risk trend is detected (i.e., a high crude anomaly risk growth rate), the frequency compensation coefficient is increased, and the second acquisition frequency is automatically increased proportionally, thereby obtaining more detailed equipment operating status data. This mechanism enables refined dynamic monitoring of the reducer status. On the one hand, it optimizes the allocation of monitoring resources through differentiated frequency strategies, ensuring comprehensive monitoring while avoiding resource waste. On the other hand, through dynamic parameter adjustment driven by risk trends, it significantly enhances the timeliness and sensitivity of anomaly detection, which can improve monitoring accuracy in the early stages of failure, trigger early warnings, and gain more intervention time for maintenance personnel, effectively avoiding serious damage to equipment, and ultimately achieving improved flexibility and reliability of predictive maintenance.

[0050] In one possible implementation, step S300 further includes:

[0051] Step S310: Configure a first preset warning indicator and a second preset warning indicator, wherein the first preset warning indicator includes a rough risk coefficient threshold and a rough risk frequency threshold, and the second preset warning indicator includes a fine risk coefficient threshold and a fine risk frequency threshold, and the rough risk coefficient threshold is smaller than the fine risk coefficient threshold, and the rough risk frequency threshold is smaller than the fine risk frequency threshold.

[0052] Specifically, it is necessary to configure the first preset warning indicator and the second preset warning indicator to build a hierarchical abnormality determination system. Among them, the first preset warning indicator serves as a preliminary screening standard, including a crude risk coefficient threshold and a crude risk frequency threshold: the crude risk coefficient threshold is set to a relatively low risk probability value (such as 0.3), aiming to cover a wider range of potential abnormalities and avoid missed detection; the crude risk frequency threshold is defined as the number of consecutive or cumulative exceeds the crude risk coefficient threshold (for example, 3 consecutive exceeds or 5 exceeds in 10 monitoring times), which is used to eliminate the interference of accidental fluctuations. The second preset warning indicator serves as a precise alarm standard, including a precise risk coefficient threshold and a precise risk frequency threshold: the precise risk coefficient threshold is higher than the crude risk coefficient threshold (such as 0.6), and a higher risk confidence level is required; the precise risk frequency threshold is also higher than the crude risk frequency threshold (such as 5 consecutive exceeds or 7 exceeds in 10 monitoring times) to ensure the rigor of abnormality determination. Through a layered design with a small coarse threshold, a large coverage range, a high fine threshold, and strict judgment standards, a progressive early warning from preliminary risk identification to precise anomaly confirmation is achieved, which not only ensures the sensitive capture of early risks, but also avoids false alarms or missed alarms caused by a single threshold, thereby improving the reliability and practicality of the detection results.

[0053] In one possible implementation, step S300 further includes:

[0054] Step S320: constructing a large anomaly detection model based on the initial monitoring indicators.

[0055] Step S330: Monitor and obtain the operating status parameter sequence within the preset time window, perform risk identification through the anomaly detection large model in turn, and output a precise anomaly risk coefficient sequence.

[0056] Specifically, a large anomaly detection model is constructed based on the aforementioned initial monitoring indicators. The specific process is as follows: First, with all the initial monitoring indicators as constraints, a plurality of sample monitoring indicator sets containing complete indicators are collected from the historical operation and maintenance monitoring records of the tool change reducer, and the corresponding sample abnormal risk coefficient sets (reflecting the failure probability under different indicator combinations) are obtained at the same time, which together constitute a training data set. Subsequently, the training data is divided into K equal parts (10≤K≤30), and K times are selected from the K parts of data using a sampling method with replacement to generate K training sets (each training set can contain repeated data to enhance the robustness of the model). For each training set, the generative adversarial network (GAN) is supervised and trained separately. Through the game process of the generator fitting the data distribution and the discriminator distinguishing between real data and generated data, the network parameters are continuously adjusted until the branch models converge, and finally a large anomaly detection model consisting of K large anomaly detection branches is obtained. Because this model covers all the initial monitoring indicators, it can perform full-dimensional feature extraction and risk analysis on the equipment operation status, providing high-resolution detection capabilities for subsequent precise abnormal risk identification.

[0057] When the crude abnormal risk coefficient sequence meets the first preset warning indicator, the operating status parameter sequence within the preset time window (such as the last 30 minutes) is first monitored and obtained. The sequence contains all the initial monitoring indicator data (covering the complete parameter system that has not been screened as high-frequency indicators). Subsequently, the parameter sequence is input into the constructed anomaly detection model for risk identification. The specific process is: first calculate the crude risk coefficient overflow mean (i.e., the average value exceeding the crude risk coefficient threshold) when the crude abnormal risk coefficient sequence meets the first preset warning indicator, and then round the formula (crude risk coefficient overflow mean ÷ historical maximum crude risk coefficient overflow mean) × K to obtain the number of branches selected P (K is the total number of branches in the large model, 10≤K≤30), and then randomly select P branches from the K anomaly detection large branches to perform parallel analysis on the input operating status parameter sequence. Each branch independently outputs a risk probability value based on the feature patterns learned from different training subsets, and finally integrates to generate a precise abnormal risk coefficient sequence. This sequence achieves high-precision quantification of equipment abnormality risks through collaborative judgment of full-indicator data input and multi-branch models. Compared with the coarse-grained detection of small models, it significantly improves the accuracy and reliability of risk identification, providing a key basis for subsequent triggering of accurate early warnings.

[0058] In one possible implementation, step S320 further includes:

[0059] Step S321: taking the several initial monitoring indicators as constraints and based on historical operation and maintenance monitoring records, collecting multiple sample monitoring indicator sets and sample abnormality risk coefficient sets.

[0060] Step S322: Use the multiple sample monitoring indicator sets and sample abnormality risk coefficient sets as training data and divide them into K equal parts. Select K times with replacement from the K data sets to obtain a first training set. Iterate and select K times to obtain K training sets, where K is an integer greater than or equal to 10 and less than or equal to 30.

[0061] Step S323: Using the K training sets, supervise the generative adversarial network and train it until convergence, harvest K large anomaly detection branches, and build a large anomaly detection model.

[0062] Specifically, with the mentioned initial monitoring indicators (covering the complete parameter system of the tool changer reducer operating status) as constraints, multiple sample monitoring indicator sets are collected from the historical operation and maintenance monitoring records of the equipment. These samples contain actual monitoring data of all initial indicators such as vibration frequency, temperature, speed, and current. At the same time, for each sample monitoring indicator set, by analyzing historical fault records, the failure probability of the tool changer reducer under this set of indicator data is statistically analyzed and set as the sample abnormality risk coefficient to form a sample abnormality risk coefficient set. The collected samples need to cover a variety of working conditions such as normal operation of the equipment, potential abnormalities, and fault states to ensure the diversity and representativeness of the training data, and provide real and comprehensive input data support for the construction of subsequent large-scale abnormality detection models, so that the model can learn the mapping relationship between different operating states and fault risks.

[0063] The collected multiple sample monitoring indicator sets and their corresponding sample anomaly risk coefficient sets are used as training data. First, these data are evenly divided into K parts (K is an integer between 10 and 30), and each data contains part of the sample and its risk coefficient. Then, the sampling method with replacement is used to randomly select K times from the K parts of data to form the first training set. Each sampling allows the same data to be repeatedly selected to increase the diversity of the training data and the robustness of the model. By iterating the above sampling process K times, K different training sets are finally obtained. Each training set contains the characteristic distribution of the original data but with different sample combinations. This bootstrap sampling method can effectively use limited historical data to generate multiple groups of statistically significant training samples, avoid the problem of model overfitting due to insufficient data volume, and provide a rich data set for the subsequent multi-branch training of the generative adversarial network, ensuring that each branch of the large anomaly detection model can learn risk characteristics of different dimensions.

[0064] A generative adversarial network (GAN) is trained independently using supervised training on each of the K training sets (K is an integer between 10 and 30). Each training set contains a set of sample monitoring indicators and a corresponding set of sample anomaly risk coefficients generated from the original data using sampling with replacement. These serve as input data to drive the GAN's generator and discriminator in a game of learning: the generator is responsible for fitting the distribution of the input data to generate simulated data that is close to real samples; the discriminator is responsible for distinguishing between real data and generated data. The network parameters are continuously adjusted using a backpropagation algorithm until the distribution of the generator's output data closely matches the distribution of real data, making it impossible for the discriminator to accurately distinguish between the two. This is considered model convergence. After K independent training cycles, K large anomaly detection branches with different feature extraction capabilities are generated. Each branch corresponds to a set of optimized network parameters, capable of capturing different risk signature patterns from the full set of initial monitoring indicators. These K branches are combined into a large anomaly detection model, which possesses multi-branch parallel analysis capabilities. By integrating multi-dimensional features, the model improves the detection accuracy of complex abnormal conditions in tool change reducers, providing a strong model support for the subsequent output of precise anomaly risk coefficients.

[0065] In one possible implementation, step S330 further includes:

[0066] Step S331: Calculate the crude risk coefficient overflow mean when the crude abnormal risk coefficient sequence meets the first preset warning indicator.

[0067] Step S332: Calculate the ratio of the crude risk coefficient overflow mean to the historical maximum crude risk coefficient overflow mean, multiply it by K and round it up to obtain the number of branch selections P.

[0068] Step S333: randomly selecting P anomaly detection large branches from the K anomaly detection large branches of the anomaly detection large model, performing risk identification on the operating state parameter sequence, and outputting a precise anomaly risk coefficient sequence.

[0069] Specifically, when the crude abnormal risk coefficient sequence meets the first preset warning indicator (that is, the crude risk coefficient exceeds the crude risk coefficient threshold and reaches the crude risk frequency threshold), all risk coefficient data points in the sequence that exceed the crude risk coefficient threshold are first screened out. For each data point that meets the conditions, the difference between it and the crude risk coefficient threshold (that is, the actual risk coefficient minus the threshold) is calculated to obtain the overflow value. Subsequently, the arithmetic mean of these overflow values is calculated to obtain the crude risk coefficient overflow mean. This mean quantifies the average degree to which the current abnormal risk exceeds the preliminary warning threshold, and provides key parameters for subsequent assessment of risk severity and dynamic adjustment of the number of branches of the fine detection model, ensuring that a more refined detection process can be triggered when the equipment risk increases.

[0070] The calculated mean crude risk coefficient overflow (i.e., the average degree to which the current crude risk coefficient exceeds the threshold) is compared with the mean maximum crude risk coefficient overflow in the historical records, and the ratio of the two is calculated to reflect the relative level of the current risk overflow. Subsequently, this ratio is multiplied by the total number of branches K (10≤K≤30) of the anomaly detection large model, and the number of selected branches P is obtained by rounding. This process normalizes historical data so that the P value can dynamically respond to the magnitude of risk changes. When the risk overflow is high, the P value increases, and more model branches participate in fine detection; conversely, the number of participating branches is reduced. Through this mechanism, computing resources can be intelligently allocated according to the severity of the risk, optimizing computing efficiency while ensuring detection accuracy, ensuring that the anomaly detection large model can achieve efficient risk identification at different risk levels.

[0071] From the K anomaly detection branches (K is an integer between 10 and 30) of the anomaly detection large model, according to the calculated number of branches P, P branches are randomly selected to participate in the precise anomaly risk identification. The selection process adopts a random strategy to ensure that the branch combination of each detection is diverse and avoid detection bias caused by fixed branches. For the operating status parameter sequence within the preset time window (including all initial monitoring indicator data), the selected P branches perform risk identification in parallel, and each branch independently outputs a risk coefficient value based on the feature pattern learned from its training subset. Finally, the arithmetic mean of the output results of the P branches is taken to generate a precise anomaly risk coefficient sequence. This multi-branch random selection and result fusion mechanism can not only improve detection efficiency through parallel computing, but also utilize the feature complementarity of different branches to improve the reliability of detection results, so that the precise anomaly risk coefficient sequence more accurately reflects the actual abnormal risk level of the equipment.

[0072] Example 2: Figure 2 This is a structural diagram of an electronic device provided in Example 2 of the present application, and is a block diagram of an exemplary electronic device suitable for implementing an embodiment of the present invention. Figure 2 The electronic device shown is only an example and should not limit the functions and scope of use of the embodiments of the present invention. Figure 2 As shown, the electronic device includes a processor 21, a memory 22, an input device 23 and an output device 24; the number of processors 21 in the electronic device can be one or more. Figure 2 Taking a processor 21 as an example, the processor 21, memory 22, input device 23 and output device 24 in the electronic device can be connected through a bus or other means. Figure 2 The bus connection is taken as an example.

[0073] In a third embodiment, the memory 22, as a computer-readable storage medium, can be used to store software programs, computer-executable programs, and modules, such as the program instructions / modules corresponding to the abnormality detection method for a tool magazine tool changer reducer in the embodiments of the present application. The processor 21 executes the software programs, instructions, and modules stored in the memory 22 to execute various functional applications and data processing of the computer device, thereby implementing the abnormality detection method for a tool magazine tool changer reducer described above.

[0074] It should be noted that the order in which the embodiments of the present application are presented is for illustrative purposes only and does not necessarily represent the superiority or inferiority of the embodiments. Furthermore, the foregoing descriptions of specific embodiments of this specification are provided. Furthermore, the processes depicted in the accompanying drawings do not necessarily require the specific order or sequential sequence shown to achieve the desired results. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0075] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present application should be included in the scope of protection of the present application.

[0076] This specification and drawings are merely illustrative of the present application and are intended to cover any and all modifications, variations, combinations, or equivalents within the scope of this application. Obviously, those skilled in the art may make various modifications and variations to this application without departing from the scope of this application. Thus, this application is intended to include such modifications and variations as fall within the scope of this application and its equivalents.

Claims

1. A method for detecting abnormality of a tool magazine tool change reducer, characterized in that: Methods include: During the machining process, the high-frequency operating state parameter set of the tool change reducer is acquired by fixed-point monitoring according to the first acquisition frequency. Risk identification is performed through a small anomaly detection model, and a rough anomaly risk coefficient sequence is output. Obtaining the operating status parameter set of the tool-changing reducer by fixed-point monitoring according to the second acquisition frequency; If the rough abnormal risk coefficient sequence meets the first preset warning indicator, the operating state parameter set is subjected to risk identification through the abnormality detection large model, and a fine abnormal risk coefficient sequence is output; If the precise abnormality risk coefficient sequence meets the second preset warning index, an abnormality warning is issued for the tool changing reducer.

2. The abnormality detection method of a tool magazine tool change reducer according to claim 1 is characterized in that: According to the first acquisition frequency fixed-point monitoring, the high-frequency operating status parameter set of the tool change reducer is obtained, including: Acquire preset operating status monitoring parameters of the tool-changing reducer, wherein the preset operating status monitoring parameters include a plurality of initial monitoring indicators; Performing correlation analysis on the several initial monitoring indicators and the abnormal state of the tool change reducer respectively, and selecting the initial monitoring indicators with the maximum correlation in a preset proportion as high-frequency monitoring indicators to obtain a high-frequency monitoring indicator set, wherein the preset proportion is less than 30%; Based on the high-frequency monitoring indicator set, a high-frequency operating status parameter set of the tool-changing reducer is acquired by fixed-point monitoring according to the first acquisition frequency.

3. The abnormality detection method of a tool magazine tool change reducer according to claim 2 is characterized in that: Risk identification is performed through a small anomaly detection model, which outputs a rough anomaly risk coefficient sequence, including: Taking the high-frequency monitoring indicator set as a constraint, multiple sample high-frequency monitoring indicator sets are collected based on the historical operation and maintenance monitoring records of the tool changer reducer, and the failure probability of the tool changer reducer under different sample high-frequency monitoring indicator set states is statistically calculated and set as the sample abnormality risk coefficient to obtain the sample abnormality risk coefficient set; Using the multiple sample high-frequency monitoring indicator sets and the sample abnormality risk coefficient sets as training data, training the generative adversarial network until convergence, and obtaining a small anomaly detection model; The high-frequency operating status parameter set sequence within the preset time window is monitored and obtained, and the anomaly detection small model is input into the fixed point to perform risk identification, and a rough anomaly risk coefficient sequence is output.

4. The abnormality detection method of a tool magazine tool change reducer according to claim 2, characterized in that: According to the second acquisition frequency fixed-point monitoring, the operating status parameter set of the tool change reducer is obtained, including: Based on the plurality of initial monitoring indicators, fixed-point monitoring is performed according to a second acquisition frequency to obtain an operating status parameter set of the tool-changing reducer, wherein the second acquisition frequency is less than the first acquisition frequency; Configure a preset time period for risk trend judgment, obtain a sequence of crude abnormal risk coefficients within the preset time period, perform crude abnormal risk trend analysis, and output a crude abnormal risk growth rate; Using 1 plus the product of the rough abnormal risk growth rate and a preset constant as the frequency compensation coefficient, wherein the preset constant is less than or equal to 15; The second acquisition frequency is periodically adjusted according to the frequency compensation coefficient.

5. The abnormality detection method for a tool magazine tool change reducer according to claim 1 is characterized in that: Configure a first preset warning indicator and a second preset warning indicator, wherein the first preset warning indicator includes a rough risk coefficient threshold and a rough risk frequency threshold, and the second preset warning indicator includes a fine risk coefficient threshold and a fine risk frequency threshold, and the rough risk coefficient threshold is smaller than the fine risk coefficient threshold, and the rough risk frequency threshold is smaller than the fine risk frequency threshold.

6. The abnormality detection method for a tool magazine tool change reducer according to claim 2, characterized in that: The risk identification of the operating status parameter set is performed through the anomaly detection large model, and a precise anomaly risk coefficient sequence is output, including: Building a large anomaly detection model based on the several initial monitoring indicators; The monitoring obtains the operating status parameter sequence within the preset time window, and then identifies risks through the anomaly detection model in turn, and outputs a precise anomaly risk coefficient sequence.

7. The abnormality detection method for a tool magazine tool change reducer according to claim 6, characterized in that: A large anomaly detection model is constructed based on the aforementioned initial monitoring indicators, including: Taking the aforementioned initial monitoring indicators as constraints and based on historical operation and maintenance monitoring records, collecting multiple sample monitoring indicator sets and sample abnormality risk coefficient sets; The multiple sample monitoring indicator sets and sample abnormality risk coefficient sets are used as training data and divided equally into K parts. K parts are selected with replacement from the K parts of the data set to obtain a first training set. K parts are selected iteratively K times to obtain K training sets, where K is an integer greater than or equal to 10 and less than or equal to 30; Using the K training sets, supervised training is performed on the generative adversarial network until convergence, and K large anomaly detection branches are obtained to form a large anomaly detection model.

8. The abnormality detection method for a tool magazine tool change reducer according to claim 7, characterized in that: Monitor and obtain the operating status parameter sequence within the preset time window, identify risks through the anomaly detection model in turn, and output a precise anomaly risk coefficient sequence, including: Calculating a crude risk coefficient overflow mean when the crude abnormal risk coefficient sequence meets a first preset warning indicator; Calculate the ratio of the crude risk coefficient overflow mean to the historical maximum crude risk coefficient overflow mean, multiply it by K and round it up to obtain the number of branches to be selected, P; Among the K anomaly detection large branches of the anomaly detection large model, P anomaly detection large branches are randomly selected to perform risk identification on the operating state parameter sequence and output a precise anomaly risk coefficient sequence.

9. An electronic device, characterized in that: The electronic device comprises: a memory for storing executable instructions; The processor is configured to implement the method for detecting an abnormality of a tool magazine tool changing reducer according to any one of claims 1 to 8 when executing the executable instructions stored in the memory.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, a method for detecting an abnormality of a tool magazine tool change reducer according to any one of claims 1 to 8 is implemented.

Images