Intelligent adaptive tool wear monitoring and compensation method and system

Abstract

The application provides a kind of intelligent adaptive tool wear monitoring and compensation method and system, wherein the method comprises: real-time acquisition of multi-modal sensor data of tool in machining process, extract and fuse to obtain comprehensive wear characteristics;Based on dynamic wear rate model, the instantaneous wear rate of the tool is calculated;Dynamic wear rate model at least with comprehensive wear characteristics and cumulative processing time as input, output instantaneous wear rate, and update the current cumulative wear of tool according to this;Based on the current cumulative wear and instantaneous wear rate, the remaining useful life of the tool is predicted and the future wear state evolution trajectory is generated;According to the current cumulative wear of tool and wear state evolution trajectory, the required tool compensation amount at future key state node is dynamically calculated, and automatic compensation operation is performed.The method provided by the application realizes real-time, accurate and forward-looking management of tool wear state, effectively avoids the decline of machining quality and non-planned shutdown caused by monitoring lag or compensation not in time.

Description

Technical Field

[0001] This disclosure generally relates to the field of CNC machine tool technology, and specifically to an intelligent adaptive tool wear monitoring and compensation method. Background Technology

[0002] As core equipment in modern manufacturing, CNC machine tools rely heavily on the performance of their cutting tools for machining accuracy and efficiency. As vulnerable components directly involved in cutting, tool wear directly impacts workpiece quality, potentially leading to workpiece scrap and equipment downtime. Therefore, real-time and accurate monitoring of tool wear and effective compensation are crucial for ensuring machining stability and improving production efficiency.

[0003] However, existing tool wear monitoring and compensation methods still have significant limitations: First, most technologies rely on a single type of sensor signal, which is poorly adaptable to complex and changing working conditions and prone to misjudgment; second, wear rate calculations mostly use static empirical models, which cannot be dynamically adjusted according to actual machining parameters, material properties, and environmental factors, resulting in insufficient prediction accuracy; third, monitoring and compensation are often disconnected, with compensation actions lagging behind wear occurrence and lacking adaptive capabilities based on forward-looking predictions. These shortcomings collectively restrict further improvements in machining accuracy and efficiency, and also make it difficult to achieve truly intelligent closed-loop control. Summary of the Invention

[0004] In view of the above-mentioned defects or deficiencies in the prior art, it is desirable to provide an intelligent adaptive tool wear monitoring and compensation method and system to solve the above problems.

[0005] The first aspect of this application provides an intelligent adaptive tool wear monitoring and compensation method, including: S1: Real-time acquisition of multimodal sensing data of the tool during the machining process, and extraction and fusion of comprehensive wear characteristics that can characterize the wear state of the tool from the multimodal sensing data; S2: Calculate the instantaneous wear rate of the tool based on the dynamic wear rate model; the dynamic wear rate model takes at least the comprehensive wear characteristics and cumulative machining time as input, outputs the instantaneous wear rate, and updates the current cumulative wear of the tool accordingly; S3: Based on the current cumulative wear amount and the instantaneous wear rate, predict the remaining service life of the tool and generate a future wear state evolution trajectory, the wear state evolution trajectory including one or more key state nodes and their expected trigger times; S4: Based on the current cumulative wear of the tool and the wear state evolution trajectory, dynamically calculate the tool compensation amount required for the future key state nodes, and automatically execute the compensation operation when the machining process reaches the corresponding node.

[0006] According to the technical solution provided in the embodiments of this application, step S1 includes: Based on the current processing conditions, at least one activation configuration is selected from multiple preset sensing modal configurations; wherein, the processing conditions include at least processing type and processing material, and different sensing modal configurations define combinations of sensing data types to be collected; Based on the activation configuration, the corresponding multimodal sensing data is collected; The collected multimodal sensing data is subjected to feature extraction and fusion to obtain the comprehensive wear characteristics.

[0007] According to the technical solution provided in the embodiments of this application, the method further includes a visual calibration step: During a set interval when the tool is in a stopped state, an image of the tool's cutting edge is acquired through a vision sensor, and the visual wear amount of the tool is obtained through image processing. The current cumulative wear is calibrated using the visual wear measurement, and the parameters in the dynamic wear model are corrected in reverse.

[0008] According to the technical solution provided in the embodiments of this application, the dynamic wear rate model in step S2 is as follows:

[0009] in, Indicates instantaneous wear rate; The reference wear rate is based on Taylor's tool life formula and is calculated according to the initially set cutting speed, feed rate, depth of cut, and the machinability parameters corresponding to the workpiece material grade. The dynamic wear rate factor is obtained through nonlinear fusion based on parameters affecting workpiece material, cutting fluid state, machine tool thermal state, wear characteristic deviation, and cutting load strength. This represents the normalized result of the overall wear characteristics. This represents the normalized result of the cumulative processing time. This represents a dimensionless function with comprehensive wear characteristics and cumulative processing time as input variables.

[0010] According to the technical solution provided in the embodiments of this application, step S3 includes: A digital twin of the cutting tool is created, the data model of which includes at least: the current cumulative wear amount, the historical fusion feature sequence, the instantaneous wear rate sequence, and the latest parameters of the dynamic wear rate model; Using the historical fusion feature sequence and instantaneous wear rate sequence stored in the digital twin as input, and utilizing the trained long short-term memory network model, the short-term predicted wear sequence within the first set time window in the future is output. Using the current cumulative wear amount in the digital twin as the initial state, and the dynamic wear model of the latest parameter set as the state transition equation, the particle filter algorithm is used and the Monte Carlo method is used to iterate forward to output a set of long-term predicted paths representing multiple possible wear paths within a future second set time window; the second set time window is longer than the first set time window. The short-term predicted wear sequence is weighted and fused with the long-term predicted path set to generate the wear state evolution trajectory.

[0011] According to the technical solution provided in the embodiments of this application, the key status node includes at least: wear warning node, size compensation node and forced tool change node; step S3 further includes synchronizing the key status node and its expected trigger time to the production management system and the machine tool CNC system.

[0012] According to the technical solution provided in the embodiments of this application, step S4 includes: Based on the wear state evolution trajectory, the dynamic compensation rate is calculated before reaching the next size compensation node; the dynamic compensation rate is characterized by the rate of change of the compensation amount required to offset the predicted wear within this time interval. Multiply the dynamic compensation rate by the estimated time to reach the compensation node of that size to obtain the baseline look-ahead compensation amount; The baseline look-ahead compensation amount is corrected based on the confidence level of the wear state evolution trajectory to generate a dynamic tool compensation amount; When the machining process reaches the expected trigger time of the dimensional compensation node, the dynamic tool compensation amount is applied to the machine tool CNC system.

[0013] According to the technical solution provided in the embodiments of this application, the collected multimodal sensing data includes processing audio signals collected through a microphone; the method further includes: Frequency domain features are extracted from the audio signal and compared in real time with a preset reference audio spectrum that characterizes the health status of the cutting tool. When an energy surge exceeding a preset threshold occurs in a specific frequency band, it is determined that the tool has experienced sudden wear, and an alarm is immediately triggered and / or an emergency shutdown operation is performed.

[0014] According to the technical solution provided in the embodiments of this application, the feature extraction and fusion of the collected multimodal sensing data is performed using a fusion algorithm based on attention weights; wherein, the attention weights are dynamically adjusted according to the current processing conditions.

[0015] A second aspect of this application provides an intelligent adaptive tool wear monitoring and compensation system, applied to the intelligent adaptive tool wear monitoring and compensation method described above, the system comprising: The data acquisition module is configured to acquire multimodal sensing data of the tool during the machining process in real time, and extract and fuse comprehensive wear characteristics that can characterize the wear state of the tool from the multimodal sensing data. A data calculation module is configured to calculate the instantaneous wear rate of the tool based on a dynamic wear rate model; the dynamic wear rate model takes at least the comprehensive wear characteristics and the cumulative machining time as input, outputs the instantaneous wear rate, and updates the current cumulative wear of the tool accordingly; The wear prediction module is configured to predict the remaining service life of the tool and generate a future wear state evolution trajectory based on the current cumulative wear amount and the instantaneous wear rate. The wear state evolution trajectory includes one or more key state nodes and their expected trigger times. The tool compensation module is configured to dynamically calculate the tool compensation amount required for future key state nodes based on the current cumulative wear of the tool and the wear state evolution trajectory, and automatically execute the compensation operation when the machining process reaches the corresponding node.

[0016] Compared with existing technologies, the advantages of this application are as follows: By constructing a complete technical closed loop integrating multimodal sensor data fusion, dynamic wear rate modeling, wear state prediction, and adaptive compensation, real-time, accurate, and forward-looking management of tool wear state is achieved; the method utilizes complementary multi-source information to improve the reliability of state perception and adaptively tracks working condition changes based on a dynamic model, thereby significantly improving the accuracy of wear prediction; furthermore, by strongly coupling the predicted wear evolution trajectory with compensation decisions, early and accurate compensation at key nodes is achieved, effectively avoiding the decline in machining quality and unplanned downtime caused by monitoring lag or untimely compensation in traditional methods. Ultimately, while improving machining accuracy and efficiency, the tool life is extended, and the intelligence and reliability of the manufacturing process are enhanced. Attached Figure Description

[0017] Other features, objects, and advantages of this application will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings: Figure 1 The flowchart of the intelligent adaptive tool wear monitoring and compensation method provided in Example 1 is shown below. Figure 2 This is a schematic diagram of the intelligent adaptive tool wear monitoring and compensation system provided in Example 4.

[0018] The reference numerals are as follows: 10, Data Acquisition Module; 20, Data Calculation Module; 30, Wear Prediction Module; 40, Tool Compensation Module. Detailed Implementation

[0019] The present application will now be described in further detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and not intended to limit it. Furthermore, it should be noted that, for ease of description, only the parts relevant to the invention are shown in the accompanying drawings.

[0020] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. This application will now be described in detail with reference to the accompanying drawings and embodiments.

[0021] Example 1 Please refer to Figure 1 This embodiment provides an intelligent adaptive tool wear monitoring and compensation method, including: S1: Real-time acquisition of multimodal sensing data of the tool during the machining process, and extraction and fusion of comprehensive wear characteristics that can characterize the wear state of the tool from the multimodal sensing data.

[0022] Specifically, the system integrates various sensors, including but not limited to power sensors, vibration sensors, and acoustic emission sensors. In step S1, the system synchronously acquires physical signals during the machining process through the integrated sensors, forming a multimodal sensing data stream. Subsequently, the system preprocesses and extracts features from these raw, high-dimensional sensing data. The extracted features are further processed through data fusion technology to ultimately generate a comprehensive index that more fully and stably characterizes the tool wear state, i.e., comprehensive wear characteristics. The fusion of multimodal data aims to overcome the shortcomings of single signal sources, such as susceptibility to interference and limited characterization capabilities.

[0023] S2: Calculate the instantaneous wear rate of the tool based on the dynamic wear rate model; the dynamic wear rate model takes at least the comprehensive wear characteristics and cumulative machining time as input, outputs the instantaneous wear rate, and updates the current cumulative wear of the tool accordingly.

[0024] Specifically, the system pre-builds and maintains a dynamic wear rate model. This model is trained using comprehensive wear characteristics and the cumulative machining time of the tool as the main inputs, and instantaneous wear rate as the output. The comprehensive wear characteristics effectively reflect the severity of the current cutting condition, while the cumulative machining time reflects the time factor of wear accumulation. In step S2, while obtaining the comprehensive wear characteristics in step S1, the cumulative machining time of the tool is also obtained. Then, the comprehensive wear characteristics and the cumulative machining time are input into the dynamic wear rate model. The model automatically calculates a dynamically changing instantaneous wear rate, which more accurately reflects the actual wear rate under the current working conditions. By integrating the instantaneous wear rate, the system updates and outputs the current cumulative wear of the tool in real time, achieving high-frequency online tracking of the tool's state.

[0025] S3: Based on the current cumulative wear amount and the instantaneous wear rate, predict the remaining service life of the tool and generate a future wear state evolution trajectory, which includes one or more key state nodes and their expected trigger times.

[0026] Specifically, in step S3, the system uses the current cumulative wear amount obtained in step S2 as a starting point and the changing trend reflected by the actual wear rate as a basis to use a prediction algorithm to predict the future development of tool wear. The prediction result not only includes an overall estimate of the remaining service life, but more importantly, it outputs a wear state evolution trajectory that changes over time.

[0027] Based on this trajectory, the system can identify key status nodes such as wear warning, size compensation, and forced tool change, and accurately predict the future trigger time of these nodes.

[0028] The wear warning node is triggered when the predicted cumulative wear reaches a preset tool life threshold (e.g., 80% of the maximum allowable wear). This node does not directly perform compensation but instead sends a warning signal to the production management system, alerting operators that the tool life is about to run out and preparations for tool replacement should be made, thus avoiding unplanned downtime. Based on the wear evolution trajectory, the system predicts that when machining subsequent workpieces (or subsequent features of the same workpiece), tool wear will cause the machined dimensions to approach the lower or upper tolerance limits. The system sets a dimensional compensation node before this dimensional deviation risk occurs, its core purpose being to provide precise triggering timing for subsequent proactive compensation. The forced tool change node is triggered when the predicted cumulative wear reaches the final tool life threshold (i.e., the maximum allowable wear). At this point, the system considers compensation insufficient to meet machining quality requirements and will forcibly pause the machining process through the machine tool's CNC system, issuing a tool change command to prevent complete tool failure from damaging the workpiece or machine tool.

[0029] Furthermore, step S3 also includes synchronizing the key status nodes and their expected trigger times to the production management system and the machine tool CNC system.

[0030] Specifically, step S3 also includes system synchronization: all the aforementioned key status nodes and their expected trigger times are synchronized in real-time to the production management system and the machine tool CNC system via a communication interface. This allows the production management system to optimize production scheduling and material preparation, while the machine tool CNC system can anticipate and prepare to respond to upcoming compensation or tool change commands, achieving closed-loop linkage and collaborative control between the detection and early warning system and the production execution unit.

[0031] S4: Based on the current cumulative wear of the tool and the wear state evolution trajectory, dynamically calculate the tool compensation amount required for the future key state nodes, and automatically execute the compensation operation when the machining process reaches the corresponding node.

[0032] Specifically, the current cumulative wear reflects the existing wear, while the wear state evolution trajectory represents the future trend. In step S4, the system comprehensively considers the current cumulative wear and the wear state evolution trajectory of the tool, and dynamically calculates the required tool compensation amount at the upcoming tool compensation node. This calculation method ensures that the compensation decision is both based on the current situation and forward-looking. Finally, when the machining timeline advances to the expected trigger time of this node, the system automatically sends the calculated compensation amount to the machine tool CNC system to complete precise compensation, thereby correcting the wear before it actually affects machining accuracy.

[0033] Further, step S1 includes: S1.1: Based on the current processing conditions, select at least one activation configuration from a plurality of preset sensing modal configurations; wherein, the processing conditions include at least processing type and processing material, and different sensing modal configurations define combinations of sensing data types to be collected.

[0034] In step S1.1, the system does not always collect data from all available sensors. Specifically, the system has a built-in configuration knowledge base, which pre-stores several sensor mode configurations. Each configuration clearly specifies a set of sensors recommended to be activated under a specific processing condition. The processing condition is defined by at least two dimensions: processing type (e.g., roughing, finishing) and workpiece material (e.g., 45 steel, 304 stainless steel).

[0035] For example, the system can be pre-configured as "Heavy-duty Roughing," which, for "roughing" type and "alloy steel" material, defines the combination of activated sensor data types as spindle power signal and X / Y / Z three-axis cutting force signal to focus on monitoring macroscopic mechanical loads. Simultaneously, another pre-configured setting is "Precision Milling," which, for "finishing" type and "aluminum alloy" material, defines the activation of vibration and acoustic emission signals to sensitively capture minute chatter and microscopic damage.

[0036] During system operation, the system parses CNC code in real time or receives instructions from the upper-level production management system, automatically identifying the current machining type and workpiece material. Subsequently, based on the identified working conditions, it queries the configuration knowledge base and selects one or more of the most matching activation configurations.

[0037] S1.2: Based on the activation configuration, collect the corresponding multimodal sensing data.

[0038] Step S1.2 is the execution phase. Specifically, once an activation configuration is selected, the system's data acquisition hardware and driver software are configured, and only the sensors specified in that configuration are activated for synchronous or asynchronous data acquisition. For sensors not included in the activation configuration, the system can put them into sleep or low-power states. This strategy effectively conserves system computing resources and energy while avoiding unnecessary data redundancy.

[0039] S1.3: Perform feature extraction and fusion on the collected multimodal sensing data to obtain the comprehensive wear characteristics.

[0040] In step S1.3, the system extracts features from the various sensor data streams collected according to the activation configuration in the time domain (such as mean, root mean square), frequency domain (such as the main frequency amplitude obtained by fast Fourier transform), and time and frequency domain (such as the energy features obtained by wavelet transform) to form the original feature set.

[0041] Subsequently, an attention-weighted fusion algorithm is employed. This algorithm assigns weights to different types of features, and these attention weights are not fixed but adjusted in real time according to the current machining conditions. For example, in the "finishing" condition, higher weights are given to high-frequency vibration features; while in the "roughing" condition, higher weights are given to spindle power features. Finally, through weighted synthesis, a low-dimensional comprehensive wear feature that optimally represents the current tool wear state is output. This method effectively improves the robustness and accuracy of the state representation.

[0042] Furthermore, the dynamic wear rate model in step S2 is as follows:

[0043] in, Indicates instantaneous wear rate; The reference wear rate is based on Taylor's tool life formula and is calculated according to the initially set cutting speed, feed rate, depth of cut, and the machinability parameters corresponding to the workpiece material grade. The dynamic wear rate factor is obtained through nonlinear fusion based on parameters affecting workpiece material, cutting fluid state, machine tool thermal state, wear characteristic deviation, and cutting load strength. This represents the normalized result of the overall wear characteristics. This represents the normalized result of the cumulative processing time. This represents a dimensionless function with comprehensive wear characteristics and cumulative processing time as input variables.

[0044] Specifically, in step S2, the system first processes the comprehensive wear characteristics and cumulative processing time output in step S1. Normalization was performed to obtain the normalized results of the comprehensive wear characteristics. Normalized results of cumulative processing time Subsequently, according to and The state-time coupling factor can then be obtained. This characterizes the coupled influence of the instantaneous health status reflected by the comprehensive wear characteristics and the cumulative effect reflected by the cumulative processing time on the wear rate.

[0045] In this embodiment, the state-time coupling factor Calculate using the following two methods: Method 1: ; in, and This represents the weighting coefficients determined using historical data; Indicates the linear deviation from the instantaneous health status. An increase indicates a deterioration in the health of the cutting tool, and this directly and proportionally increases the wear rate. This describes the saturation effect of cumulative effects, depicting normal, time-related progressive wear. This approach is suitable for conditions where acute wear factors (condition) and chronic wear factors (time) are relatively independent, and their effects can be linearly superimposed. For example, in relatively stable conditions and routine machining of materials.

[0046] Option 2: ; in, and This represents the weighting coefficients determined using historical data; This represents the nonlinear amplification effect of health status on wear rate, when When the wear and tear is very large (in very poor condition), even if the time is short, the wear and tear will accelerate rapidly. This represents the nonlinear accelerating effect of cumulative time on wear rate; as time progresses, the tool matrix fatigues, and at this point, the same... This can lead to more severe wear consequences. This method is suitable for working conditions where the state and time factors of wear are strongly coupled and have positive feedback. For example, when machining difficult-to-cut materials (such as high-temperature alloys), initial micro-chipping can cause a sharp increase in cutting force and temperature, thus drastically accelerating the subsequent wear process. The exponential effect will be very obvious; when the tool enters the rapid wear zone, the deterioration of the tool condition and the cumulative effect promote each other, forming a vicious cycle.

[0047] The choice between the two methods mentioned above depends on the specific processing conditions; and , and , The two types of weighting coefficients are also related to the processing conditions and are obtained by retrieving the weighting configuration mapping table. Based on historical processing data, a weighting configuration mapping table is pre-established through experimental calibration. The system queries this mapping table to obtain the corresponding weighting coefficient values ​​according to the current processing type and workpiece material. It should be noted that each processing condition in the weighting configuration mapping table corresponds to only one type of weighting coefficient. , or , .

[0048] Therefore, in step S2, before calculating the instantaneous wear rate, the following steps are also included: The weight configuration mapping table is called according to the processing type and processing material to obtain the corresponding weight coefficient value, and the weight type is determined at the same time; Obtain the corresponding state-duration coupling factor based on the weight type. ; Based on state-time coupling factor The dynamic wear rate model is modified.

[0049] In addition to calculating the instantaneous wear rate before The determination of this also requires analysis of the dynamic wear rate factor. To determine, specifically including: The dynamic wear influence parameters are obtained, including workpiece material influence parameters, cutting fluid state influence parameters, machine tool thermal state influence parameters, wear characteristic deviation parameters, and cutting load strength parameters. The dynamic wear influence parameters are input into the nonlinear fusion model to obtain the dynamic wear rate factor. .

[0050] Specifically, the material influence parameter is the workpiece material batch coefficient: derived by back-calculating the deviation between real-time cutting force and theoretical value to compensate for micro-fluid fluctuations in material hardness; the cutting fluid state influence parameter is the cutting fluid efficiency index: quantifying the decline in cooling and lubrication effect by comprehensively considering the flow rate, temperature, and concentration monitoring values ​​of the cutting fluid; the machine tool thermal state influence parameter is the machine tool thermal drift: the impact of tool-workpiece relative position changes on the cutting state estimated by a thermal deformation model based on temperature sensor data at key machine tool points; the wear characteristic deviation parameter is the real-time characteristic deviation degree: the cosine similarity between the multi-modal fusion feature of the current period and the tool health state benchmark feature, directly reflecting the instantaneous health degradation of the tool; the cutting load strength parameter represents the relative strength of cutting parameters: the normalized load strength of the current actual cutting parameters (linear velocity, feed, depth of cut) relative to the tool design recommended parameters. In this embodiment, the nonlinear fusion model is a small neural network model; it uses the workpiece material influence parameter, cutting fluid state influence parameter, machine tool thermal state influence parameter, wear characteristic deviation parameter, and cutting load strength parameter as inputs, and a dynamic wear rate factor as input. The output is obtained through training.

[0051] Further, step S3 includes: S3.1: Create a digital twin of the tool, whose data model includes at least: current cumulative wear, historical fusion feature sequence, instantaneous wear rate sequence, and the latest parameters of the dynamic wear rate model.

[0052] In step S3.1, the system creates and maintains a virtual digital twin for each physical tool. This twin is a dynamically updated data model, and at the time of execution in step S3, it contains at least the following fields: Current cumulative wear: from the latest output of step S2; Historical fusion feature sequence: a set of comprehensive wear feature values ​​stored in chronological order and periodically generated by step S1; Real wear rate sequence: A set of instantaneous wear rate values ​​stored in chronological order and periodically generated by step S2; Latest parameters for the dynamic wear rate model: including calculation of the dynamic wear rate factor Nonlinear models and state-time coupling factors The set of weight coefficients.

[0053] S3.2: Using the historical fusion feature sequence and instantaneous wear rate sequence stored in the digital twin as input, and utilizing the trained long short-term memory network model, output the short-term predicted wear sequence within the first set time window in the future.

[0054] In step S3.2, the system takes the historical fusion feature sequence and instantaneous wear rate sequence of a certain length stored in the digital twin as input and feeds them into a pre-trained Long Short-Term Memory (LSTM) network model. This LSTM model learns the temporal dependencies in the historical data and outputs a short-term predicted wear sequence within a predetermined time window (e.g., the next 10 sampling periods). Example: Assuming the current cumulative wear is 0.23 mm, the LSTM model outputs a sequence like this: [0.231, 0.233, 0.236, 0.239, 0.243, 0.248, 0.254, 0.261, 0.269, 0.278] mm. This sequence reflects the deterministic trend of wear change in the short term.

[0055] S3.3: Using the current cumulative wear amount in the digital twin as the initial state, and the dynamic wear model of the latest parameter set as the state transition equation, the particle filtering algorithm is used and the Monte Carlo method is used to iterate forward to output a set of long-term predicted paths representing multiple possible wear paths within the second set time window in the future; the second set time window is longer than the first set time window.

[0056] In step S3.3, the system uses the current cumulative wear in the digital twin as the initial state and a dynamic wear rate model incorporating the latest parameters as the state transition equation. Subsequently, a particle filter algorithm is used to perform a large number (e.g., N=1000 times) of random forward iterative simulations via the Monte Carlo method. Each iteration generates a possible future wear path considering model and operating condition uncertainties, ultimately outputting a long-term predicted path set containing thousands of possible paths. The second set of time windows covered is much longer than the first set of time windows (e.g., covering the next 200 sampling periods or until the wear threshold is reached). Example: This set is not a single value, but a two-dimensional array. At a certain future moment... The wear predictions for these 1,000 paths form a distribution, with the mean forming a "central trend line" and the standard deviation describing the range of uncertainty.

[0057] S3.4: The short-term predicted wear sequence and the long-term predicted path set are weighted and fused to generate the wear state evolution trajectory.

[0058] In step 3.4, the system performs a weighted fusion of the short-term predicted wear sequence (high-precision trend) output from S3.2 and the long-term predicted path set (probability distribution) output from S3.3. At any future time... Predicted value of the comprehensive wear state evolution trajectory Calculated by the following formula: ; in, This indicates the short-term predicted wear sequence at time [time]. The value; Represents the set of long-term predicted paths at time [time]. The statistical mean; The dynamic fusion weight is a weight that varies with time. A function that changes, with values ​​ranging from [0,1].

[0059] The dynamic fusion weight The method for determining it is as follows:

[0060] in, Indicates the current moment. This represents the decay time constant, whose value is set based on an assessment of the LSTM model's recent historical prediction accuracy. When the LSTM model has a small recent prediction error (high accuracy), The setting is too large, making Slower decay means shorter-term predictions have a higher weight over a longer period; conversely, slower decay means shorter-term predictions have a lower weight. More timely predictions rely on long-term projections.

[0061] Further, step S4 includes: S4.1: Based on the wear state evolution trajectory, calculate the dynamic compensation rate before reaching the next size compensation node; the dynamic compensation rate is characterized by the rate of change of the compensation amount required to offset the predicted wear within this time interval.

[0062] In step S4.1, based on the wear state evolution trajectory, the system first determines the compensation interval, starting from the current moment and ending at the expected trigger time of the next size compensation node. Next, the predicted wear amount at the current moment is read from the wear state evolution trajectory. And the predicted wear at the next size compensation node. Next, calculate the dynamic compensation rate. The calculation formula is as follows:

[0063] in, This represents the predicted wear rate. The wear-compensation mapping coefficient is a constant pre-calibrated experimentally, used to convert the predicted rate of change of wear into a compensation rate (e.g., mm / min) in the machine tool coordinate system. This coefficient establishes the physical relationship between the amount of wear and the actual dimensional deviation.

[0064] S4.2: Multiply the dynamic compensation rate by the estimated time to reach the compensation node of that size to obtain the baseline look-ahead compensation amount.

[0065] In step S4.2, the system performs the following calculations to obtain the baseline look-ahead compensation amount. : ; The upcoming dynamic compensation rate Compensation range Multiplying these values ​​yields a total baseline compensation amount necessary to offset the predicted wear within that range. This value represents the total compensation theoretically required if the prediction were perfectly accurate.

[0066] S4.3: Based on the confidence level of the wear state evolution trajectory, the baseline look-ahead compensation amount is corrected to generate a dynamic tool compensation amount.

[0067] In step S4.3, the baseline compensation amount is safely corrected based on the prediction confidence level. Specifically, the system first calls the confidence level of the current wear state evolution trajectory generated during the prediction process in step S3. Evaluation value. This value is a number between 0 and 1, calculated based on the recent performance of the prediction model (such as the average deviation between predicted and actual values). The baseline compensation is then adjusted using the following formula:

[0068] in, Indicates the dynamic tool compensation amount. Confidence weighting coefficient , is a preset parameter that determines the degree to which the confidence level affects the final compensation amount. When the confidence level approaches 1, the system fully trusts the prediction and executes the full baseline compensation amount; when the confidence level approaches 0, the system has doubts about the prediction and executes a discounted, more conservative compensation amount to avoid overcompensation when the prediction is inaccurate.

[0069] S4.4: When the machining process reaches the expected trigger time of the dimension compensation node, the dynamic tool compensation amount is applied to the machine tool CNC system.

[0070] In step S4.4, the system continuously monitors the machining process. When the real-time clock reaches the expected trigger time of the next dimensional compensation node, it automatically applies the calculated dynamic tool compensation amount. The value is sent to the machine tool's CNC system via the machine tool communication interface. The machine tool's CNC system then writes the value into the corresponding tool radius or length compensation register, so that it takes effect in real time in subsequent machining code, completing a proactive adaptive compensation operation.

[0071] Example 2 Based on Embodiment 1 above, this embodiment provides another intelligent adaptive tool wear monitoring and compensation method. The same content as Embodiment 1 will not be repeated here, except that the method further includes a visual calibration step. During a set interval when the tool is in a stopped state, an image of the tool's cutting edge is acquired through a vision sensor, and the visual wear amount of the tool is obtained through image processing. The current cumulative wear is calibrated using the visual wear measurement, and the parameters in the dynamic wear model are corrected in reverse.

[0072] Specifically, in this embodiment, the system presets one or more machining intervals (such as after each workpiece is completed, after every 30 minutes of continuous machining, or before each tool change). When the CNC program is paused or the tool returns to a safe position, the system automatically triggers the vision acquisition process. The vision sensor is installed inside the machine tool or near the tool magazine, ensuring that its optical axis is aligned with the cutting edge area of ​​the tool. Optionally, the vision sensor is an industrial CCD camera with a telecentric lens and a ring light source.

[0073] The system first performs grayscale conversion, filtering and noise reduction, and contrast enhancement on the acquired raw image. Then, it uses an edge detection algorithm (such as the Canny operator) to extract the contour line of the tool cutting edge, and combines Hough transform or template matching technology to locate the main cutting edge and the secondary cutting edge. By comparing the current cutting edge contour with a preset new tool reference contour at the pixel level, it identifies the contour missing or deformed areas caused by wear, calculates the geometric features of the wear area as the main indicator of visual wear, and then obtains the visual wear amount.

[0074] The system compares the visual wear amount obtained from image processing with the current cumulative wear amount estimated by the dynamic wear rate model. If the deviation exceeds a preset tolerance range, the model is determined to have cumulative errors and needs calibration. If the deviation is greater than a set threshold, the visual wear amount is directly used as the new current cumulative wear amount benchmark, and the cumulative wear amount state in the model is reset. If the deviation is less than or equal to the set threshold, a first-order low-pass filter or Kalman filter is used to weight and fuse the visual wear amount with the current cumulative wear amount estimated by the model, generating a more reliable calibrated cumulative wear amount. Example formula:

[0075] in, Indicates the amount of wear after calibration. Indicates visual wear and tear. This represents the current cumulative wear estimated by the model. This represents the trust weight, which can be dynamically adjusted based on the historical accuracy of the vision system.

[0076] Subsequently, based on the cumulative wear amount after visual calibration, the system reverse-engineers and corrects the baseline wear rate in the dynamic wear rate model. Dynamic wear rate factor and state-time coupling factor Weighting coefficients in , , and .

[0077] Example 3 Based on Embodiment 1 above, this embodiment provides another intelligent adaptive tool wear monitoring and compensation method. The same content as Embodiment 1 will not be repeated here; the difference lies in: The acquired multimodal sensing data includes processing audio signals acquired via a microphone; the method further includes: Frequency domain features are extracted from the audio signal and compared in real time with a preset reference audio spectrum that characterizes the health status of the cutting tool. When an energy surge exceeding a preset threshold occurs in a specific frequency band, it is determined that the tool has experienced sudden wear, and an alarm is immediately triggered and / or an emergency shutdown operation is performed.

[0078] Specifically, in this embodiment, an industrial microphone is deployed on the machine tool, serving as a fixed data source for the multimodal sensing system. The raw audio signal it collects, along with data streams from vibration sensors, current sensors, etc., collectively constitutes multimodal sensing data for tool condition monitoring. After acquiring the raw audio signal, the system processes it and compares it in real-time with a stored reference audio spectrum. When the energy of any key frequency band rises sharply within a short period (e.g., from a few milliseconds to hundreds of milliseconds) and exceeds a preset threshold, typically corresponding to minor chipping, breakage, or severe material adhesion at the tool edge, the system immediately determines that the tool has experienced sudden wear and responds accordingly. The response includes: sending a high-priority alarm message to the operator interface and production management system, displaying the message "Sudden Tool Wear Alarm," accompanied by audible and visual cues; and when the confidence level is extremely high or the process requirements are extremely stringent, the system can automatically send an emergency stop command to the machine tool's CNC system to immediately halt the machining process, preventing the fault from escalating and protecting the workpiece and machine tool.

[0079] Example 4 Based on Embodiment 1 above, this embodiment provides an intelligent adaptive tool wear monitoring and compensation system, applied to the intelligent adaptive tool wear monitoring and compensation method as described in Embodiment 1. The system includes: The data acquisition module 10 is configured to acquire multimodal sensing data of the tool during the machining process in real time, and extract and fuse comprehensive wear characteristics that can characterize the wear state of the tool from the multimodal sensing data. The data calculation module 20 is configured to calculate the instantaneous wear rate of the tool based on a dynamic wear rate model; the dynamic wear rate model takes at least the comprehensive wear characteristics and the cumulative machining time as input, outputs the instantaneous wear rate, and updates the current cumulative wear amount of the tool accordingly; Wear prediction module 30 is configured to predict the remaining service life of the tool and generate a future wear state evolution trajectory based on the current cumulative wear amount and the instantaneous wear rate. The wear state evolution trajectory includes one or more key state nodes and their expected trigger times. The tool compensation module 40 is configured to dynamically calculate the tool compensation amount required for future key state nodes based on the current cumulative wear amount of the tool and the wear state evolution trajectory, and automatically perform the compensation operation when the machining process reaches the corresponding node.

[0080] The above description is merely a preferred embodiment of this application and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of the invention involved in this application is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the inventive concept. For example, technical solutions formed by substituting the above features with (but not limited to) technical features with similar functions disclosed in this application.

Claims

1. A method for intelligent adaptive tool wear monitoring and compensation, characterized in that, include: S1: Real-time acquisition of multimodal sensing data of the tool during the machining process, and extraction and fusion of comprehensive wear characteristics that can characterize the wear state of the tool from the multimodal sensing data; S2: Calculate the instantaneous wear rate of the tool based on the dynamic wear rate model; The dynamic wear rate model takes at least the comprehensive wear characteristics and cumulative machining time as input, outputs the instantaneous wear rate, and updates the current cumulative wear of the tool accordingly; S3: Based on the current cumulative wear amount and the instantaneous wear rate, predict the remaining service life of the tool and generate a future wear state evolution trajectory, the wear state evolution trajectory including one or more key state nodes and their expected trigger times; S4: Based on the current cumulative wear of the tool and the wear state evolution trajectory, dynamically calculate the tool compensation amount required for the key state nodes in the future, and automatically execute the compensation operation when the machining process reaches the corresponding node; The dynamic wear rate model in step S2 is as follows: in, Indicates instantaneous wear rate; The reference wear rate is based on Taylor's tool life formula and is calculated according to the initially set cutting speed, feed rate, depth of cut, and the machinability parameters corresponding to the workpiece material grade. The dynamic wear rate factor is obtained through nonlinear fusion based on parameters affecting workpiece material, cutting fluid state, machine tool thermal state, wear characteristic deviation, and cutting load strength. This represents the normalized result of the overall wear characteristics. This represents the normalized result of the cumulative processing time. This represents a dimensionless function with comprehensive wear characteristics and cumulative processing time as input variables.

2. The intelligent adaptive tool wear monitoring and compensation method according to claim 1, characterized in that, Step S1 includes: Based on the current processing conditions, at least one activation configuration is selected from multiple preset sensing modal configurations; wherein, the processing conditions include at least processing type and processing material, and different sensing modal configurations define combinations of sensing data types to be collected; Based on the activation configuration, the corresponding multimodal sensing data is collected; The collected multimodal sensing data is subjected to feature extraction and fusion to obtain the comprehensive wear characteristics.

3. The intelligent adaptive tool wear monitoring and compensation method according to claim 2, characterized in that, The method also includes a visual calibration step: During a set interval when the tool is in a stopped state, an image of the tool's cutting edge is acquired through a vision sensor, and the visual wear amount of the tool is obtained through image processing. The current cumulative wear is calibrated using the visual wear measurement, and the parameters in the dynamic wear model are corrected in reverse.

4. The intelligent adaptive tool wear monitoring and compensation method according to claim 3, characterized in that, Step S3 includes: A digital twin of the cutting tool is created, the data model of which includes at least: the current cumulative wear amount, the historical fusion feature sequence, the instantaneous wear rate sequence, and the latest parameters of the dynamic wear rate model; Using the historical fusion feature sequence and instantaneous wear rate sequence stored in the digital twin as input, and utilizing the trained long short-term memory network model, the short-term predicted wear sequence within the first set time window in the future is output. Using the current cumulative wear amount in the digital twin as the initial state, and the dynamic wear model of the latest parameter set as the state transition equation, the particle filter algorithm is used and the Monte Carlo method is used to iterate forward to output a set of long-term predicted paths representing multiple possible wear paths within a future second set time window; the second set time window is longer than the first set time window. The short-term predicted wear sequence is weighted and fused with the long-term predicted path set to generate the wear state evolution trajectory.

5. The intelligent adaptive tool wear monitoring and compensation method according to claim 4, characterized in that, The critical status nodes include at least: wear warning node, size compensation node, and forced tool change node; step S3 also includes synchronizing the critical status nodes and their expected trigger times to the production management system and the machine tool CNC system.

6. The intelligent adaptive tool wear monitoring and compensation method according to claim 5, characterized in that, Step S4 includes: Based on the wear state evolution trajectory, the dynamic compensation rate is calculated before reaching the next size compensation node; the dynamic compensation rate is characterized by the rate of change of the compensation amount required to offset the predicted wear within this time interval. Multiply the dynamic compensation rate by the estimated time to reach the compensation node of that size to obtain the baseline look-ahead compensation amount; The baseline look-ahead compensation amount is corrected based on the confidence level of the wear state evolution trajectory to generate a dynamic tool compensation amount; When the machining process reaches the expected trigger time of the dimensional compensation node, the dynamic tool compensation amount is applied to the machine tool CNC system.

7. The intelligent adaptive tool wear monitoring and compensation method according to claim 6, characterized in that, The acquired multimodal sensing data includes processing audio signals acquired via a microphone; the method further includes: Frequency domain features are extracted from the audio signal and compared in real time with a preset reference audio spectrum that characterizes the health status of the cutting tool. When an energy surge exceeding a preset threshold occurs in a specific frequency band, it is determined that the tool has experienced sudden wear, and an alarm is immediately triggered and / or an emergency shutdown operation is performed.

8. The intelligent adaptive tool wear monitoring and compensation method according to claim 7, characterized in that, The feature extraction and fusion of the collected multimodal sensing data are performed using an attention weight-based fusion algorithm; wherein the attention weight is dynamically adjusted according to the current processing conditions.

9. An intelligent adaptive tool wear monitoring and compensation system, characterized in that, The system, applied to the intelligent adaptive tool wear monitoring and compensation method as described in any one of claims 1-8, comprises: The data acquisition module (10) is configured to acquire multimodal sensing data of the tool during the machining process in real time, and extract and fuse comprehensive wear characteristics that can characterize the wear state of the tool from the multimodal sensing data. The data calculation module (20) is configured to calculate the instantaneous wear rate of the tool based on a dynamic wear rate model; the dynamic wear rate model takes at least the comprehensive wear characteristics and the cumulative machining time as input, outputs the instantaneous wear rate, and updates the current cumulative wear of the tool accordingly; Wear prediction module (30), the wear prediction module (30) is configured to predict the remaining service life of the tool and generate a future wear state evolution trajectory based on the current cumulative wear amount and the instantaneous wear rate, the wear state evolution trajectory including one or more key state nodes and their expected trigger time; The tool compensation module (40) is configured to dynamically calculate the tool compensation amount required for the future key state nodes based on the current cumulative wear amount of the tool and the wear state evolution trajectory, and automatically perform the compensation operation when the machining process reaches the corresponding node.

Images