Embedded monitoring method for cutter wear of numerical control machine tool

Abstract

The invention relates to the technical field of state monitoring of numerical control machine tools, in particular to an embedded monitoring method for cutter wear of a numerical control machine tool. The method comprises the following steps: performing embedded monitoring and synchronous multi-modal data acquisition and preprocessing on a tool in a machining process to form a real-time working condition data frame queue; cutting effective interval data interception is carried out on the real-time working condition data frame queue to obtain an effective cutting data sequence; performing current and cutter feeding rate discretization processing on the effective cutting data sequence to form an energy-effect reference fingerprint; extracting an expected healthy feed rate from the energy-effect reference fingerprint; and performing performance deviation absolute amount calculation according to the expected healthy feed rate, and performing cutting performance attenuation analysis to obtain the cutting performance attenuation degree. According to the method, a complete process from detection to parameter adaptive closed-loop control is realized on a resource-limited embedded system, so that the service life of a cutter is prolonged, the machining quality is guaranteed, and the production efficiency is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of numerical control machine tool state monitoring, and in particular to an embedded monitoring method for tool wear of a numerical control machine tool. BACKGROUND

[0002] Traditional tool monitoring adopts a fixed threshold monitoring method, which faces a serious problem of low sensitivity. These methods usually monitor based on the absolute threshold of current or vibration, and only trigger an alarm when the parameter exceeds the preset value. Since tool wear is a gradual process, this monitoring mechanism hardly detects the slow performance decline, and only alarms when the tool is severely worn or even broken. At this time, the workpiece has been scrapped, causing material waste and production interruption. More seriously, the "normal value" under different materials and different machining conditions varies greatly, leading to a dilemma in setting the fixed threshold: setting too high causes serious missed alarms, and setting too low causes frequent false alarms.

[0003] Advanced intelligent monitoring algorithms face great challenges in resource-constrained embedded environments. Advanced monitoring methods such as machine learning and deep neural networks require powerful computing power and large memory resources, while industrial embedded controllers usually use low-power microcontrollers with only a few hundred KB of RAM and limited Flash storage, making it difficult to support complex algorithms. Existing solutions either upload data to the cloud for analysis, which causes bandwidth pressure, poor real-time performance, relies on network connection and has data security concerns; or greatly simplify the algorithm, sacrificing monitoring accuracy and reliability, forming a typical "cloud intelligence and edge intelligence" dilemma.

[0004] Multi-modal data fusion and closed-loop control implementation face serious technical barriers. Existing monitoring systems are mostly single-parameter monitoring (such as monitoring current or vibration only), which cannot fully capture the complex wear characteristics of tools; and mostly open-loop passive monitoring, which can only issue alarms and cannot actively adjust parameters, requiring operators to manually intervene to respond to abnormalities. In addition, due to the lack of cross-validation mechanism between parameters, false alarms and missed alarms are serious, and the credibility and practicality of the monitoring system are greatly compromised. Most importantly, the closed-loop control from abnormal detection to parameter adjustment requires complex decision logic, which is extremely difficult to implement in a resource-constrained embedded environment.

[0005] In summary, the existing technology has the problems of insufficient monitoring sensitivity, conflict between algorithm complexity and embedded resources, and lack of reliable multi-modal fusion and closed-loop control capabilities, which need to be solved. SUMMARY

[0006] Therefore, it is necessary to provide an embedded monitoring method for tool wear of a numerical control machine tool to solve at least one of the above technical problems.

[0007] To achieve the above object, an embedded monitoring method for tool wear of numerical control machine tools comprises the following steps: Step S1: The tool in the machining process is monitored, multi-modal data acquisition and preprocessing are synchronized, and a real-time working condition data frame queue is formed. The effective cutting data sequence is obtained by cutting effective interval data interception on the real-time working condition data frame queue. The energy-efficiency reference fingerprint is formed by discrete processing of current and tool feed rate on the effective cutting data sequence. Step S2: The expected health feed rate is extracted from the energy-efficiency reference fingerprint. The performance deviation absolute amount is calculated according to the expected health feed rate, and the cutting performance attenuation is analyzed to obtain the cutting performance attenuation degree. Step S3: The vibration morphological abnormality degree is calculated based on the real-time working condition data frame queue to obtain the vibration morphological abnormality degree. The tool wear trend time series analysis is carried out based on the cutting performance attenuation degree to obtain the wear dynamic trend value. The tool wear is determined based on the wear dynamic trend value and the vibration morphological abnormality degree, and the machining quality is optimized to generate an optimization strategy type. The feed rate dynamic adjustment coefficient is obtained by calculating the feed rate adjustment according to the optimization strategy type. Step S4: The feed rate dynamic adjustment coefficient is written into the CNC controller of the numerical control machine tool through the preset communication interface to perform machining parameter closed-loop adjustment processing, and the machining parameter closed-loop adjustment instruction is generated.

[0008] The present application solves many drawbacks of the prior art through a series of closely coupled innovative methods. First, the present application discards the low sensitivity defect of traditional fixed threshold monitoring, by learning and constructing a personalized energy-efficiency benchmark fingerprint for each new core component in step S1, and calculating the "cutting performance attenuation degree" in real time in step S2, it can capture the early performance decline caused by the gradual wear of the component with very high sensitivity, thereby realizing reliable early warning, effectively avoiding the workpiece scrap and production interruption caused by sudden failure of the component, where energy-efficiency is the corresponding relationship between "energy consumption" and "cutting efficiency" in the numerical control machine tool machining process (where "energy consumption" is represented by spindle current and effective cutting power in the cutting process, and "cutting efficiency" is represented by tool feed rate. Secondly, the present application ingeniously solves the problem of running complex algorithms in a resource-limited embedded environment. In the entire embodiment, for example, DMA hardware acceleration and lock-free queues are used in step S1, kurtosis analysis is used instead of complex spectrum analysis in step S3, and a lookup table-based expert decision matrix is used instead of a resource-intensive AI model, these lightweight algorithm designs optimized for embedded environments ensure that the entire monitoring and decision-making process can run efficiently and in real time on low-cost, low-power microcontrollers without relying on expensive cloud computing. Finally, the present application greatly improves the reliability and practical value of monitoring, by innovatively combining energy-efficiency models (reflecting gradual wear) and "vibration morphology" analysis (capturing impact failures) in step S3, the false positive rate is significantly reduced. More importantly, through intelligent decision-making in step S3 and parameter closed-loop injection in step S4, the present application builds a complete high-speed closed loop from "detection-diagnosis-decision-control" in an embedded local environment, enabling the system to automatically adjust operating parameters based on the real-time health status of the component, extending the service life of the component, maximizing its residual value while maximizing the quality of the machining process, completely changing the limitations of traditional monitoring systems that can only passively alarm, and giving the equipment real autonomous optimization capabilities. BRIEF DESCRIPTION OF DRAWINGS

[0009] Figure 1 Figure 1 is a step flowchart of an embedded monitoring method for tool wear of a numerical control machine tool. DETAILED DESCRIPTION

[0010] The technical method of the present application will be described below in conjunction with the drawings. Obviously, the described embodiments are part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of the present application.

[0011] Furthermore, the accompanying drawings are included to provide a further understanding of the present application, and are incorporated in and constitute a part of this specification. The drawings illustrate embodiments of the present application and, together with the description, serve to explain the principles of the present application. In the drawings:

[0012] It should be understood that, although the terms "first" can be used herein to describe various elements, these elements should not be limited by these terms. The terms are only used to distinguish one element from another. The term "and / or" as used herein includes any and all combinations of one or more of the associated associated items.

[0013] To achieve the above object, there is provided Figure 1 The present application provides an embedded monitoring method for tool wear of a numerical control machine tool, comprising the following steps: Step S1: performing embedded monitoring on the tool in the machining process, synchronously collecting and preprocessing multi-modal data to form a real-time working condition data frame queue; performing data interception on the effective cutting interval of the real-time working condition data frame queue to obtain an effective cutting data sequence; and performing discretization processing on the current and tool feed rate of the effective cutting data sequence to form an energy-efficiency reference fingerprint; In this embodiment, first, multi-modal data such as spindle current, speed, feed rate and vibration are synchronously collected through embedded real-time interruption, and "instantaneous effective cutting power" is calculated in the interruption. Then, these data are packaged into "real-time working condition data frames" and stored in a queue. Next, in the background analysis task, through statistical methods, idle and non-steady cutting data are automatically identified and removed to obtain pure "effective cutting data sequence". Finally, the "effective cutting power" in the sequence is discretely mapped, and the "actual feed rate" corresponding to each interval is statistically averaged, thereby constructing and solidifying an energy-efficiency reference fingerprint lookup table describing the standard efficiency of a new tool at different powers.

[0014] Step S2: extracting an expected healthy feed rate from the energy-efficiency reference fingerprint; performing performance deviation absolute amount calculation according to the expected healthy feed rate, and performing cutting performance attenuation analysis to obtain a cutting performance attenuation degree; In this embodiment, this step is performed in a background analysis task. For each real-time working condition data frame in the active cutting state, the task indexes the "instantaneous effective cutting power" in the solidified energy-efficiency benchmark fingerprint, and quickly queries to obtain an "expected health feed rate". Then, the expected value is subtracted from the "actual feed rate" contained in the data frame to obtain an "absolute deviation of feed rate" with physical units. To eliminate the inconsistency of the evaluation scale under different working conditions, the absolute deviation is further normalized and smoothed through an IIR low-pass filter to finally obtain a stable, dimensionless "cutting performance attenuation degree".

[0015] Step S3: Perform vibration morphology abnormality calculation on the real-time working condition data frame queue to obtain a vibration morphology abnormality; perform tool wear trend time series analysis based on the cutting performance attenuation degree to obtain a wear dynamic trend value; perform tool wear determination based on the wear dynamic trend value and the vibration morphology abnormality, and perform machining quality optimization to generate an optimization strategy type; perform feed rate adjustment calculation according to the optimization strategy type to obtain a feed rate dynamic adjustment coefficient; In this embodiment, this step is performed in a background decision task. First, the task calculates two core diagnostic indicators in parallel: one is to calculate the "wear dynamic trend value" reflecting the speed of change by comparing the short-term and long-term means of the "cutting performance attenuation degree"; the other is to obtain a "vibration morphology abnormality" reflecting the risk of machining surface quality by performing time domain statistics (such as calculating kurtosis) on the original data of the vibration signal without complex spectrum analysis. Then, a lightweight embedded expert system matches the data in three dimensions of "cutting performance attenuation degree", "wear dynamic trend value" and "vibration morphology abnormality" with an externally set "quality assurance priority" for multi-objective decision making, and selects an "optimization strategy type" from the pre-set strategy library that best fits the current scenario. Finally, according to the selected strategy type, through a multi-factor coupled calculation formula, the current attenuation degree and future deterioration trend are considered comprehensively, and safety boundary and change slope are constrained, to solve the final "feed rate dynamic adjustment coefficient".

[0016] Step S4: Write the feed rate dynamic adjustment coefficient into the CNC controller of the numerical control machine tool through the pre-set communication interface to perform machining parameter closed-loop adjustment processing, and generate machining parameter closed-loop adjustment instructions; In this embodiment, this step is performed in a separate background communication task. The task first identifies the brand and model of the currently connected CNC controller through automatic detection and polling mechanism, and loads the matching communication protocol, register address, data format and legal value range from the internal "protocol adaptation template library". Then, according to the loaded template, it performs boundary clipping, format conversion and byte sequence adjustment on the "feed rate dynamic adjustment coefficient" calculated in the previous step to generate a "formatted control value" that the controller can directly parse. Then, it encapsulates the value into a "parameter control message" that meets the specific industrial bus protocol, with checksum and sequence number. Finally, the message is sent to the CNC controller through the corresponding physical interface, and the parameter is verified for successful writing through reading the reply or reverse query, thereby completing a complete and reliable local closed-loop control action from monitoring, diagnosis to autonomous optimization.

[0017] Preferably, in step S1, the tool in the machining process is monitored in embedded mode, and synchronous multi-modal data acquisition and preprocessing include: A trigger signal for starting a machining task is obtained, embedded real-time interrupt configuration and resource reservation processing are performed, and an embedded sampling interrupt environment is formed, wherein the embedded sampling interrupt environment includes the real-time state of the configured hardware timer, DMA channel and statically pre-allocated buffer; Based on the embedded sampling interrupt environment, multi-channel data is efficiently and synchronously collected to obtain an original sensor data set; The original sensor data set is subjected to embedded adaptive calibration of baseline parameters to obtain a compensated no-load current baseline; Based on the original sensor data set and the compensated no-load current baseline, embedded optimized power real-time calculation is performed to obtain an effective cutting power value; The effective cutting power value and the original sensor data set are subjected to embedded real-time data queue management to form a real-time working condition data frame queue.

[0018] In an embodiment, the microcontroller of the embedded device configures its on-chip hardware timer (TIM2) upon receiving a rising edge level signal from the I / O port of the numerical control system as a trigger signal for starting a machining task, sets its reload value to 499 and its pre-frequency value to 71, thereby generating an update interrupt with a period of 1 millisecond. The priority of the update interrupt is set to the highest priority 0 in the system. At the same time, the direct memory access (DMA) controller is configured to set its data transmission direction to from peripheral to memory, the transmission mode to circular mode, and a block of memory with a size of 256 bytes is statically pre-allocated as a data receiving buffer. The configured hardware timer, DMA channel and statically pre-allocated buffer collectively constitute the embedded sampling interrupt environment.

[0019] In each interrupt service routine (ISR) generated by a hardware timer (TIM2) with a period of 1 millisecond, a DMA controller is automatically triggered to initiate a sequence acquisition of a multi-channel analog-to-digital converter (ADC) without intervention of the microcontroller core. The sequence acquisition synchronously acquires analog voltage values of a spindle motor current sensor and X, Y, Z axis analog voltage values of a three-axis acceleration sensor tightly connected to a tool holder in strict accordance with a preset channel order. At almost the same time, the microcontroller core sends a remote transmission request (RTR) frame with an ID of 0x101 to the servo drivers of the spindle and feed shaft through a controller area network (CAN) bus to read the original digital values of the current rotation speed and the current feed rate in the internal registers of the servo drivers. All the acquired original sensor readings jointly constitute the original sensor dataset.

[0020] In a learning phase of device initialization, a separate low-priority task is used to count the number of original sensor datasets in 1000 consecutive original sensor datasets acquired, in which the spindle motor current value is lower than a preset rough threshold (for example, 2 amperes). By constructing a current value frequency histogram including 20 intervals, the interval with the highest frequency is determined, and the arithmetic mean of all current values in the interval is calculated as the "initial no-load current mean value". During normal operation, the statistical process is repeated every 100 machining cycles, and a first-order IIR low-pass filter is used to update the "initial no-load current mean value" in the following manner: new mean value = 0.1 x current statistical mean value + 0.9 x old mean value. The mean value after the adaptive update is the compensated no-load current baseline.

[0021] According to the original sensor dataset and the compensated no-load current baseline, embedded optimization of power real-time calculation is performed to obtain an effective cutting power value. This step decouples and purifies multiple physical quantities into a more essential indicator through a lightweight calculation in the interrupt service routine. Specifically, after each original sensor dataset is acquired, a subtraction and multiplication operation is immediately performed, which is calculated in the following manner: ; wherein, is the effective cutting power value, is the instantaneous current value in the original sensor dataset, is the compensated no-load current baseline, is a bus read or preset DC bus voltage value (for example, 310 volts), is a preset motor efficiency coefficient (for example, 0.85).

[0022] The effective cutting power value obtained by this calculation has stripped away the effects of no-load power consumption and rotation speed changes, and better represents the net power actually used for material removal compared to the original current value.

[0023] The effective cutting power value calculated in the previous step, along with the raw values ​​of the actual feed rate and triaxial acceleration from the original sensor dataset, and a timestamp provided by the counter value of the hardware timer (TIM2), are encapsulated into a fixed-size 24-byte structure as a "real-time operating condition data frame". Subsequently, this data frame is stored in a pre-allocated 4-kilobyte circular buffer through a single memory copy operation, and the write pointer is atomically moved. This circular buffer constitutes the real-time operating condition data frame queue, serving as an asynchronous communication hub between high-priority interrupt tasks and medium-priority analysis tasks. Its lock-free design ensures efficient data transmission and the stability of the real-time system.

[0024] Preferably, step S1, which involves extracting the effective cutting interval data from the real-time operating condition data frame queue, includes: The no-load current feature is extracted from the real-time operating condition data frame queue to obtain the no-load current reference value; The cutting interval is initially identified by using the no-load current reference value to obtain the original cutting interval set from the real-time operating condition data frame queue. The original cutting interval set is corrected for working condition change boundaries to obtain the cutting interval boundaries; Based on the boundaries of the cutting intervals, the steady-state cutting segments of the real-time working condition data frame queue are identified to obtain the set of steady-state cutting intervals. By traversing the set of steady-state cutting intervals, effective data extraction and merging are performed on the real-time working condition data frame queue to obtain an effective cutting data sequence.

[0025] In another embodiment, at the start of each processing cycle, the analysis task continuously retrieves 1000 data frames from the real-time operating condition data frame queue and extracts the instantaneous current values ​​of the spindle motor contained therein. A frequency histogram of current values ​​covering the range of 0 to 3 amperes with a resolution of 0.1 amperes is constructed to determine the interval with the highest frequency. The arithmetic mean and standard deviation of all current values ​​within this interval are calculated, and these two statistics are used as the reference value for the no-load current.

[0026] Next, the cutting interval is initially identified in the real-time operating condition data frame queue using the no-load current reference value. This process utilizes statistical principles to generate a dynamic threshold to distinguish between cutting and non-cutting operations. Specifically, a cutting initiation threshold is set, with the value of: ;in, For the cutting start threshold value for determining whether the tool enters the effective cutting state, μ is the average value in the no-load current reference value, and σ is the standard deviation in the no-load current reference value. The data frames in the real-time working condition data frame queue are checked one by one, if the instantaneous current value of a data frame first exceeds the cutting start threshold value, the timestamp of the data frame is marked as the start of a segment of cutting; if the instantaneous current value of a subsequent data frame first falls below the cutting start threshold value, the timestamp of the data frame is marked as the end of the segment of cutting. All the marked start-end timestamp pairs together constitute the original cutting interval set.

[0027] To eliminate the interference of transient impact caused by spindle acceleration or deceleration or workpiece entering or leaving the cutting area on the recognition result, the preliminary recognized interval boundaries are corrected. Specifically, each interval in the original cutting interval set is traversed, and the data of 50 milliseconds at the start and end stages of the interval is discarded. For example, for an interval from time point t_1 to t_2, the new interval after correction is [t_1+50ms, t_2-50ms]. This operation aims to eliminate the unstable state at the start and end of each cutting segment, and ensure that the subsequent analysis is based on stable cutting data. All the new intervals after correction together constitute the cutting interval boundary.

[0028] According to the cutting interval boundary, the real-time working condition data frame queue is subjected to stable cutting segment recognition. This step aims to further purify the data and identify the intervals that are truly used for stable machining. Specifically, within each interval defined by the cutting interval boundary, the average value and the standard deviation of the "effective cutting power" values of all the data frames in the interval are calculated. If the ratio of the power standard deviation to the average value (i.e. the coefficient of variation) of a certain interval is less than a preset stable discrimination coefficient (for example, 0.15), it is determined that the interval is a stable cutting segment. All the intervals determined to be stable cutting segments together constitute the cutting stable interval set.

[0029] According to the time range recorded by the cutting stable interval set, the analysis task extracts only the data frames with timestamps falling within these stable intervals from the vast real-time working condition data frame queue. Then, the data frames extracted from all the stable cutting segments are merged into a single, continuous data sequence in chronological order, which is the effective cutting data sequence. The effective cutting data sequence provides a pure data source for subsequent construction of high-precision energy-efficiency benchmark fingerprints.

[0030] Preferably, the discretization processing of the effective cutting data sequence in step S1 includes: power domain discretization interval mapping of the effective cutting data sequence to obtain a discrete interval rate set; interval feature value statistics of the discrete interval rate set to form interval feature values; The discrete interval rate set and interval characteristic value are solidified as the energy-efficiency benchmark fingerprint.

[0031] In an embodiment, the valid cutting data sequence is mapped to power domain discrete intervals. The continuous, multi-dimensional machining state data is converted to a structured, discrete model for efficient query. Specifically, the analysis task first traverses the valid cutting data sequence to determine the maximum value P_max and the minimum value P_min of all "valid cutting power" values in the sequence. Then, the power range [P_min, P_max] is evenly divided into N (e.g. N = 50) non-overlapping power discrete intervals. Each interval is uniquely identified by its starting power value. Next, each data frame in the valid cutting data sequence is traversed again, and according to its "valid cutting power" value, it is assigned to the corresponding power discrete interval, and only the "actual feed rate" value in the data frame is added to a dynamic array associated with the interval. Finally, a set of N dynamic arrays is formed, which is the discrete interval rate set.

[0032] The analysis task traverses the N dynamic arrays in the discrete interval rate set. For each non-empty dynamic array, the task calculates the arithmetic mean of all "actual feed rate" values in the array. This arithmetic mean is considered as the standard performance that a healthy tool can achieve at that particular power discrete interval. The arithmetic means calculated from all N dynamic arrays collectively form the interval characteristic value.

[0033] The discrete interval rate set and interval characteristic value are solidified as the energy-efficiency benchmark fingerprint. Specifically, the analysis task creates a Look-Up Table (LUT) structure containing N entries. Each entry in the LUT uses the starting power value of its corresponding power discrete interval as the "Key", and the "interval characteristic value" (i.e. the rate average) calculated in the previous step as the "Value". After the construction, the complete LUT data structure is written to the flash memory or EEPROM of the embedded device through the standard non-volatile memory write interface for persistent storage. This solidified LUT is the energy-efficiency benchmark fingerprint, which provides an efficient and fast query benchmark for the subsequent online performance quantification step.

[0034] Preferably, step S2 comprises: extracting the current spindle motor instantaneous current and actual feed rate from the real-time working condition data frame queue to construct an instantaneous working condition data pair; extracting the current instantaneous current value from the instantaneous working condition data pair, and using it as an index to perform benchmark rate extraction on the energy-efficiency benchmark fingerprint to obtain the expected healthy feed rate; According to the instantaneous working condition data pair and the expected healthy feed rate, a performance deviation absolute value is calculated to obtain a feed rate absolute deviation; Using the feed rate absolute deviation and the expected healthy feed rate, attenuation degree normalization and smoothing filtering are performed to obtain a cutting performance attenuation degree.

[0035] In an embodiment, the current effective cutting power and the actual feed rate are extracted from the real-time working condition data frame queue. Specifically, the analysis task continuously takes out a data frame from the head of the real-time working condition data frame queue during normal machining of the machine tool. If the data frame is determined to be in the effective cutting interval, the “effective cutting power” value and the “actual feed rate” value are extracted therefrom, and the two values are combined into a data pair as the instantaneous working condition data pair.

[0036] The effective cutting power value is extracted from the instantaneous working condition data pair and used as an index to extract the reference rate from the energy-efficiency reference fingerprint. The real-time working condition is compared with the historical health model to establish a reference for performance evaluation. Specifically, the analysis task extracts the “effective cutting power” value from the instantaneous working condition data pair. Then, using the power value as an index, the power discrete interval containing the power value is quickly located in the energy-efficiency reference fingerprint (a lookup table) stored in the non-volatile memory through a binary search algorithm. Then, the “value” corresponding to the interval is directly read, that is, the standard healthy feed rate, as the expected healthy feed rate.

[0037] Then, according to the instantaneous working condition data pair and the expected healthy feed rate, a performance deviation absolute value is calculated. Specifically, the “actual feed rate” value is extracted from the instantaneous working condition data pair. Then, a subtraction operation is performed: the expected healthy feed rate obtained in the previous step is subtracted from the actual feed rate. If the operation result is positive, the absolute value thereof is the feed rate absolute deviation; if it is negative or zero, the feed rate absolute deviation is zero. The deviation value is directly quantified in millimeters per minute, which quantifies how much slower the current tool is than the healthy tool under the same power consumption.

[0038] To eliminate the inconsistency of the evaluation scale under different cutting powers, normalization processing is performed. Specifically, a division operation is performed: the feed rate absolute deviation is divided by the expected healthy feed rate as a reference to obtain a dimensionless percentage value. To suppress signal noise, the percentage value is further input into a first-order IIR low-pass filter for smoothing processing, and the update method is as follows: ; wherein, is the updated smoothed value, is the current calculated normalized percentage value, is the last smoothed value, is a smoothing coefficient (for example, 0.2).

[0039] The value after this normalization and smoothing filtering is the cutting performance attenuation degree, which will be updated to a globally shared memory area for use by other tasks.

[0040] Preferably, the performance deviation absolute amount calculation of the expected healthy feed rate according to the instantaneous working condition data pair comprises: The working condition data synchronous pairing of the instantaneous working condition data pair and the expected healthy feed rate to obtain a feed rate pairing value; The absolute performance difference calculation according to the feed rate pairing value to obtain an original feed difference value; The working condition adaptability processing according to the original feed difference value and the instantaneous working condition data pair to obtain a working condition calibration difference value; The transient fluctuation suppression processing using the working condition calibration difference value to obtain a steady-state performance difference value; The physical meaning mapping verification of the steady-state performance difference value and the instantaneous working condition data pair to obtain a feed rate absolute deviation.

[0041] In another embodiment, since the instantaneous working condition data pair and the expected healthy feed rate are both derived from the same "effective cutting power" value at the same time, this step combines these two data which are strictly synchronized in time and working condition reference. Specifically, a temporary data structure containing two members is constructed, one member is the "actual feed rate" extracted from the instantaneous working condition data pair, and the other member is the "expected healthy feed rate", and this data structure is the feed rate pairing value.

[0042] Then, the absolute performance difference calculation according to the feed rate pairing value. Specifically, a subtraction operation is performed: subtract the "actual feed rate" member from the "expected healthy feed rate" member in the feed rate pairing value. The result of this operation is the original feed difference value, which directly reflects the instantaneous gap between theoretical performance and actual performance.

[0043] Then, the working condition adaptability processing according to the original feed difference value and the instantaneous working condition data pair. This step dynamically adjusts the tolerance of the deviation according to the fineness of the machining. Specifically, the task extracts the "effective cutting power" value from the instantaneous working condition data pair. If this power value is lower than a preset "finishing power threshold" (for example, 500 watts), it is determined that the current working condition is finishing, and the original feed difference value is multiplied by a "finishing sensitivity coefficient" greater than 1 (for example, 1.5); if the power value is higher than the threshold, it is determined to be rough machining, and the original feed difference value is kept unchanged. The difference value after this multiplication operation adjustment is the working condition calibration difference value.

[0044] Afterwards, the transient fluctuation suppression processing is performed by using the working condition calibration difference. In order to filter out the sharp jump of single data point caused by material hard point or measurement noise, the continuously calculated working condition calibration difference is smoothed. Specifically, the arithmetic mean value of the latest five working condition calibration differences is calculated by using a sliding window containing five data points. The mean value is the steady state performance difference, which can represent the average performance deviation condition in a recent period of time more than the single original difference.

[0045] Finally, the steady state performance difference and the instantaneous working condition data pair are mapped and verified in physical meaning. Since the tool wear is a one-way performance degradation process, only the performance loss has a clear physical meaning. Specifically, the steady state performance difference is judged once: if the value is positive, it means that there is a performance loss, and it is directly used as the absolute deviation of the feed rate; if the value is negative or zero, it means that the performance has not decreased or even slightly exceeded, and it does not have a physical meaning of wear, so the absolute deviation of the feed rate is forced to be zero. This operation ensures that the final output deviation value is always non-negative, and only has a non-zero value when the performance decays.

[0046] Preferably, the vibration morphological abnormality degree calculation on the real-time working condition data frame queue in step S3 includes: segmenting and buffering the vibration data of the real-time working condition data frame queue to obtain a vibration segmented data block; extracting low-computational-statistical features from the vibration segmented data block to obtain a vibration statistical feature set; detecting morphological feature mutations from the vibration statistical feature set to obtain a morphological mutation feature; performing impact feature enhancement processing according to the morphological mutation feature and the vibration statistical feature set to obtain an enhanced impact feature; calculating the vibration morphological abnormality degree based on the enhanced impact feature.

[0047] In an embodiment, the vibration data of the real-time working condition data frame queue is segmented and buffered. Specifically, the analysis task continuously takes N (for example, N=256) data frames from the real-time working condition data frame queue which are judged to be in the effective cutting interval. Then, only the three-axis acceleration original values are extracted from these data frames, and the vector module is calculated to form a resultant acceleration time sequence containing N points. The sequence is stored in a dedicated static array as the vibration segmented data block.

[0048] Then, low-computational statistical feature extraction is performed on the vibration segmented data block. This step synchronously calculates multiple statistical moments in one pass to achieve the highest computational efficiency. Specifically, by single-passing the vibration segmented data block, a set of accumulator variables is used to synchronously calculate the first four central moments of the data block. Then, based on these central moments, two core morphological features, namely kurtosis and skewness, are calculated by the following formulas: ; The formula for calculating skewness is: ; wherein is the kurtosis, is the skewness, is the fourth central moment, is the third central moment, is the standard deviation.

[0049] The calculated kurtosis value and skewness value together constitute the vibration statistical feature set. Then, morphological feature mutation detection is performed on the vibration statistical feature set. Specifically, the analysis task maintains a history buffer that stores the last 10 calculated vibration statistical feature sets. By calculating the difference between the current kurtosis value and the average of the kurtosis values in the history buffer, the kurtosis change amount is obtained. If the kurtosis change amount exceeds a preset mutation detection threshold (for example, the absolute value change exceeds 2.0), it is determined that a morphological mutation has occurred. The judgment result (a Boolean value, true or false) is the morphological mutation feature.

[0050] Subsequently, impact feature enhancement processing is performed according to the morphological mutation feature and the vibration statistical feature set. This step aims to amplify the impact signal indicating early failure such as chipping. Specifically, if the morphological mutation feature is true, the current kurtosis value in the vibration statistical feature set is multiplied by an impact enhancement coefficient greater than 1 (for example, 1.5); if the morphological mutation feature is false, the current kurtosis value remains unchanged. The kurtosis value after this conditional multiplication operation is the enhanced impact feature.

[0051] Finally, the vibration morphological abnormality degree is calculated based on the enhanced impact feature. Specifically, the enhanced impact feature is compared with the "healthy vibration kurtosis benchmark" value recorded in the first step of the learning phase. By calculating the absolute value of the difference between the two and normalizing it, a dimensionless percentage value is finally obtained, which is the vibration morphological abnormality degree. It comprehensively reflects the severity of the current vibration waveform deviating from the healthy state due to impact events.

[0052] Preferably, the tool wear determination based on the wear dynamic trend value and the vibration morphological abnormality degree in step S3 and the machining quality optimization include: The wear level is determined by partitioning the wear dynamic trend value, and a tool wear level indicator is obtained. According to the tool wear level indicator and the vibration form abnormality and the cutting performance attenuation, the machining quality influence prediction is performed, and a quality guarantee priority is obtained. According to the quality guarantee priority and the preset optimization strategy library, a multi-objective strategy matching is performed, and an optimization strategy type is generated.

[0053] In an embodiment, the wear level is determined by partitioning the wear dynamic trend value. Specifically, the decision task compares the wear dynamic trend value read from the shared memory area with a "wear grading threshold table" fixed in the memory. The table divides the wear dynamic trend value into four levels: less than 0.05 is level 0 (normal); [0.05, 0.1) is level 1 (mild wear); [0.1, 0.2) is level 2 (moderate wear); and greater than or equal to 0.2 is level 3 (severe wear). By one interval judgment, the current belonging level is determined, and the code (0, 1, 2, 3) of the level is the tool wear level indicator.

[0054] Then, according to the tool wear level indicator and the vibration form abnormality and the cutting performance attenuation, the machining quality influence prediction is performed. This step fuses multiple heterogeneous diagnostic indicators to evaluate the risk to the final product quality. Specifically, an initial quality influence score is calculated by the following formula: ; wherein, is the quality influence score, is the cutting performance attenuation, is the vibration form abnormality, and is a preset weight coefficient (for example, =0.6, =0.4), and are influence coefficients queried from the built-in lookup table according to the current machining material and process type. Then, an over-level judgment rule is added: if the tool wear level indicator is 1 (mild), but the vibration form abnormality instantaneously exceeds a high-risk threshold (for example, 0.8), the quality influence score is forcibly set to the highest score 10. The final 0 to 10 score obtained is the quality guarantee priority.

[0055] Finally, the multi-objective strategy matching is performed according to the quality assurance priority and the preset optimization strategy library. This step maps the multi-dimensional input to a specific action strategy through a lightweight expert decision system. Specifically, the tool wear level indicator, the quality assurance priority, and a "trend state" discretized from the wear dynamic trend value are combined into a decision vector. Then, the vector is used to query a "N-dimensional expert decision matrix" solidified in the memory. Each cell of the matrix has pre-stored an encoding representing a specific optimization strategy (for example, 0x01 represents the efficiency priority strategy, 0x02 represents the quality priority strategy, 0x03 represents the life extension strategy, and 0x04 represents the protection strategy). For example, when the decision vector is [level 2, priority > 7, trend deterioration], the result of querying the matrix is 0x04. This encoding queried from the matrix is the optimization strategy type.

[0056] Preferably, the feed rate adjustment calculation according to the optimization strategy type in step S3 includes: calculating a basic adjustment coefficient according to the optimization strategy type and the cutting performance attenuation degree; analyzing the size and direction of the wear dynamic trend value to perform trend response compensation on the basic adjustment coefficient to form a trend-compensated coefficient; obtaining the current process parameter information, and performing process adaptability adjustment on the trend-compensated coefficient to obtain a process-adjusted coefficient; performing boundary constraint processing on the process-adjusted coefficient to obtain a constraint-adjusted coefficient; performing time domain smoothing processing on the constraint-adjusted coefficient to obtain a feed rate dynamic adjustment coefficient.

[0057] In an embodiment, the basic adjustment coefficient is calculated according to the optimization strategy type and the cutting performance attenuation degree. Specifically, the decision task uses the optimization strategy type code to query the "basic attenuation response coefficient" k_base corresponding to the strategy from a built-in "strategy calculation parameter table". Then, a preliminary adjustment value reflecting the current wear state is calculated through a linear formula, which is calculated as follows: ; wherein, k_base is the basic adjustment coefficient, is the cutting performance attenuation degree.

[0058] Then, the size and direction of the wear dynamic trend value are analyzed to perform trend response compensation on the basic adjustment coefficient. This step introduces foresight into the control system and responds to the deterioration speed of the wear. Specifically, the "trend response coefficient" k_trend corresponding to the current strategy is queried from the "strategy calculation parameter table". Then, the trend influence is coupled in through the following formula: ; wherein, is the trend-compensated coefficient, is the wear rate extracted from the wear dynamic trend value. This calculation achieves a predictive compensation for future wear. Then, the current process parameter information is obtained, and the process adaptability is adjusted in combination with the trend-compensated coefficient. Specifically, the decision task reads the current machining condition identified in the first step from the shared memory area, such as “rough machining” or “finishing”. If it is currently in the “rough machining” condition, the trend-compensated coefficient is multiplied by a “rough machining tolerance coefficient” (for example, 1.1) greater than 1 to allow higher efficiency to be maintained in the non-critical machining stage; if it is in the “finishing” condition, it remains unchanged. The value after this machining condition adaptability multiplication operation is the process-adjusted coefficient.

[0059] After that, the process-adjusted coefficient is used for boundary constraint processing. To ensure safety, the calculated coefficient value is clamped in this step. Specifically, the process-adjusted coefficient is compared with a preset “absolute safety range” (for example, [0.5, 1.0]). If the calculated value is lower than 0.5, it is forcibly set to 0.5; if it is higher than 1.0, it is forcibly set to 1.0. This operation ensures that the output instruction is always within the safe boundary acceptable by the machine tool. The value after this boundary constraint processing is the constraint-adjusted tuning coefficient.

[0060] Finally, the constraint-adjusted tuning coefficient is subjected to time domain smoothing processing. To avoid the impact of dramatic changes in control instructions on the machine tool, the output value is subjected to slope limitation in this step. Specifically, the difference between the constraint-adjusted tuning coefficient and the last final output value is calculated. If the absolute value of the difference exceeds a preset “maximum change slope” (for example, no more than 0.05 change per 100 milliseconds), the output value of this time will be limited so that the change amount is exactly equal to the maximum slope. The value after this smoothing processing is the feed rate dynamic tuning coefficient, which will be used as the final control quantity and transmitted to the next step.

[0061] Preferably, step S4 comprises: detecting the state of the preset communication interface to obtain communication link state information; performing CNC controller compatibility analysis according to the communication link state information and the feed rate dynamic tuning coefficient to obtain controller parameter mapping information; performing instruction data format conversion on the feed rate dynamic tuning coefficient according to the controller parameter mapping information to obtain a formatted control value; constructing a parameter control message based on the formatted control value and the communication link state information; performing parameter injection execution and verification on the parameter control message to obtain a machining parameter closed-loop adjustment instruction.

[0062] In an embodiment, the state of the preset communication interface is detected. Specifically, the communication task detects the physical connection state in sequence according to a preset priority list (for example: EtherCAT, Profinet, Modbus TCP, CANopen) when starting. A protocol-specific link probe packet (such as an EtherCAT link state query frame) is sent, and a valid response is waited for. The first interface that obtains a valid response is determined as the current active interface. The interface type, measured response time, and the like are recorded as the communication link state information.

[0063] Then, the CNC controller compatibility analysis is performed according to the communication link state information and the feed rate dynamic adjustment coefficient. This step realizes plug and play of the device and self-adaptation to different brand controllers. Specifically, after the active interface is determined, the communication task sends a standard manufacturer and device ID reading instruction through the interface. According to the returned ID information, a template matched with the ID is queried and loaded from a “controller protocol adaptation template library” solidified in the memory. The template precisely defines the register address, data type, legal value range, and byte sequence corresponding to the “feed rate adjustment” function. These parameters queried from the template together constitute the controller parameter mapping information.

[0064] Then, the instruction data format conversion of the feed rate dynamic adjustment coefficient is performed according to the controller parameter mapping information. Specifically, the communication task converts the logical percentage form of the feed rate dynamic adjustment coefficient into a physical data format that can be recognized by the target controller. For example, if the data type defined in the controller parameter mapping information is a 16-bit unsigned integer and the range is 0-1000 corresponding to 0-100%, then the multiplication and rounding operation is performed: code value = round(feed rate dynamic adjustment coefficient x 1000). If the byte sequence defined is the big-endian mode, then the high and low bytes of the 16-bit code value are also swapped. The value after the conversion and byte sequence adjustment is the formatted control value.

[0065] After that, the parameter control packet is constructed based on the formatted control value and the communication link state information. Specifically, the communication task calls the corresponding packet construction function according to the protocol type determined in the communication link state information. For example, for the Modbus TCP protocol, the task constructs a standard Modbus packet, in which the function code is set to 0x06 (write single register), the register address is set to the address obtained from the controller parameter mapping information, and the data written is the formatted control value generated in the previous step. The transaction identifier of the packet is set to a self-incrementing sequence number, and the packet header is calculated and attached. The byte stream constructed is the parameter control packet.

[0066] Finally, the parameter injection execution and verification is performed on the parameter control packet. Specifically, the communication task sends the parameter control packet out through the selected physical interface (e.g. Ethernet port) and starts a timeout timer (e.g. 50 ms). If a correct response packet is received from the controller before the timeout, the write is considered successful. To verify, the task sends a read request for the same register address immediately after and compares the returned value with the sent value. If they are the same, the parameter is confirmed to be effective. A complete sending and verification record, including target address, sent data, time stamp and success status, constitutes the processing parameter closed-loop adjustment instruction.

[0067] Therefore, from any point of view, the embodiments should be considered as exemplary and non-limiting, the scope of the application being defined by the appended claims and not by the above description, all the variations falling within the meaning and the scope of the equivalent elements of the application file being intended to be included in the application.

[0068] The foregoing is considered as merely illustrative of the principles of the application, and the application is to be regarded as not limited to such embodiments, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. An embedded monitoring method for tool wear in a numerical control machine tool, characterized in that, The method comprises the following steps: Step S1: Embedded monitoring of the tool during machining, synchronous multi-modal data acquisition and preprocessing, forming a real-time working condition data frame queue; Cutting effective interval data interception is performed on the real-time working condition data frame queue to obtain an effective cutting data sequence; Discretization processing of current and tool feed rate is performed on the effective cutting data sequence to form an energy-efficiency reference fingerprint; Step S2: Extracting the expected healthy feed rate from the energy-efficiency reference fingerprint; Performance deviation absolute value calculation is performed according to the expected healthy feed rate, and cutting performance attenuation analysis is performed to obtain the cutting performance attenuation degree; Step S3: Vibration morphological abnormality degree calculation is performed on the real-time working condition data frame queue to obtain the vibration morphological abnormality degree; Based on the cutting performance attenuation degree, the wear trend time series analysis of the tool is performed to obtain the wear dynamic trend value; Based on the wear dynamic trend value and the vibration morphological abnormality degree, the tool wear is determined, and the machining quality is optimized to generate an optimization strategy type; according to the optimization strategy type, the feed rate adjustment calculation is performed to obtain the feed rate dynamic adjustment coefficient; Step S4: Through the preset communication interface, the feed rate dynamic adjustment coefficient is written into the CNC controller of the numerical control machine tool for machining parameter closed-loop adjustment processing to generate machining parameter closed-loop adjustment instructions.

2. The embedded monitoring method for tool wear of a numerical control machine tool according to claim 1, characterized in that, In step S1, the embedded monitoring of the tool during machining, synchronous multi-modal data acquisition and preprocessing comprises: A trigger signal for starting the machining task is obtained, embedded real-time interrupt configuration and resource reservation processing are performed, and an embedded sampling interrupt environment is formed, wherein the embedded sampling interrupt environment contains the real-time state of the configured hardware timer, DMA channel and statically pre-allocated buffer; Based on the embedded sampling interrupt environment, multi-channel data is efficiently and synchronously collected to obtain an original sensor data set; Baseline parameter embedded adaptive calibration is performed on the original sensor data set to obtain a compensated no-load current baseline; Based on the original sensor data set and the compensated no-load current baseline, embedded optimized power real-time calculation is performed to obtain an effective cutting power value; Embedded real-time data queue management is performed on the effective cutting power value and the original sensor data set to form a real-time working condition data frame queue.

3. The method for embedded monitoring of tool wear for a numerically controlled machine tool according to claim 1, characterized in that, In step S1, the cutting effective interval data interception on the real-time working condition data frame queue comprises: No-load current feature extraction is performed on the real-time working condition data frame queue to obtain a no-load current reference value; The real-time working condition data frame queue is preliminarily identified by using the no-load current reference value to obtain an original cutting interval set; The original cutting interval set is subjected to working condition change boundary correction to obtain a cutting interval boundary; The real-time working condition data frame queue is subjected to steady cutting section identification according to the cutting interval boundary to obtain a cutting steady interval set; The real-time working condition data frame queue is subjected to effective data extraction and merging by traversing the cutting steady interval set to obtain an effective cutting data sequence.

4. The method for embedded monitoring of a CNC machine tool tool wear according to claim 1, characterized in that, In step S1, the discretization processing of current and tool feed rate on the effective cutting data sequence comprises: Power domain discretization interval mapping is performed on the effective cutting data sequence to obtain a discrete interval rate set; Interval characteristic value statistics are performed on the discrete interval rate set to form an interval characteristic value; The discrete interval rate set and the interval characteristic value are solidified as energy-efficiency benchmark fingerprints.

5. The method for embedded monitoring of a CNC machine tool tool wear according to claim 1, characterized in that, Step S2 includes: extracting the current spindle motor instantaneous current and the actual feed rate from the real-time working condition data frame queue, constructing an instantaneous working condition data pair; extracting the current instantaneous current value from the instantaneous working condition data pair, and taking it as an index to extract the benchmark rate from the energy-efficiency benchmark fingerprint, obtaining the expected healthy feed rate; calculating the absolute amount of performance deviation according to the instantaneous working condition data pair and the expected healthy feed rate, obtaining the absolute deviation of the feed rate; normalizing and smoothing the attenuation degree using the absolute deviation of the feed rate and the expected healthy feed rate, obtaining the cutting performance attenuation degree.

6. The embedded monitoring method for tool wear of a numerical control machine tool according to claim 5, characterized in that, The absolute amount of performance deviation is calculated according to the instantaneous working condition data pair and the expected healthy feed rate, including: synchronizing and pairing the instantaneous working condition data pair and the expected healthy feed rate to obtain a feed rate pairing value; calculating the absolute performance difference value according to the feed rate pairing value to obtain an original feed difference value; processing the working condition adaptability according to the original feed difference value and the instantaneous working condition data pair to obtain a working condition calibration difference value; performing transient fluctuation suppression processing using the working condition calibration difference value to obtain a steady-state performance difference value; mapping and verifying the physical meaning of the steady-state performance difference value and the instantaneous working condition data pair to obtain the absolute deviation of the feed rate.

7. The method for embedded monitoring of a CNC machine tool tool wear according to claim 1, characterized in that, The vibration morphological abnormality degree calculation of the real-time working condition data frame queue in step S3 includes: segmenting and buffering the vibration data of the real-time working condition data frame queue to obtain vibration segmented data blocks; extracting low-computational-statistical-features from the vibration segmented data blocks to obtain a vibration statistical feature set; detecting morphological feature mutations in the vibration statistical feature set to obtain a morphological mutation feature; performing impact feature enhancement processing according to the morphological mutation feature and the vibration statistical feature set to obtain an enhanced impact feature; calculating the vibration morphological abnormality degree based on the enhanced impact feature.

8. The method for embedded monitoring of a CNC machine tool tool wear according to claim 1, characterized in that, The tool wear determination based on the wear dynamic trend value and the vibration morphological abnormality degree in step S3, and the machining quality optimization include: determining the tool wear grade according to the wear dynamic trend value to obtain a tool wear grade indication; performing machining quality impact prediction according to the tool wear grade indication, the vibration morphological abnormality degree, and the cutting performance attenuation degree to obtain a quality assurance priority; performing multi-objective strategy matching according to the quality assurance priority and a pre-set optimization strategy library to generate an optimization strategy type.

9. The method for embedded monitoring of a CNC machine tool tool wear according to claim 1, characterized in that, The feed rate adjustment calculation according to the optimization strategy type in step S3 includes: calculating a basic adjustment coefficient according to the optimization strategy type and the cutting performance attenuation degree; analyzing the size and direction of the wear dynamic trend value to perform trend response compensation on the basic adjustment coefficient to form a trend-compensated coefficient; obtaining the current process parameter information, and performing process adaptability adjustment combined with the trend-compensated coefficient to obtain a process-adjusted coefficient; performing boundary constraint processing using the process-adjusted coefficient to obtain a constraint-adjusted adjustment coefficient; performing time-domain smoothing processing on the constraint-adjusted adjustment coefficient to obtain a feed rate dynamic adjustment coefficient.

10. The method for embedded monitoring of a CNC machine tool tool wear according to claim 1, characterized in that, Step S4 includes: detecting the state of the pre-set communication interface to obtain communication link state information; According to the communication link state information and the feed rate dynamic tuning coefficient, CNC controller compatibility analysis is performed to obtain controller parameter mapping information; According to the controller parameter mapping information, the feed rate dynamic tuning coefficient is subjected to command data format conversion to obtain formatted control values; Based on the formatted control values and the communication link state information, parameter control messages are constructed; Parameter injection execution and verification are performed on the parameter control messages to obtain machining parameter closed-loop adjustment instructions.

Images