An adaptive control method and system for a batch tool machine

Abstract

The application relates to the technical field of precision hardware machining, and discloses a self-adaptive control method and system for a batch head machining machine tool, which comprises the following steps: acquiring process state characteristics in a machining cycle; based on the process state characteristics, extracting a trend reference component and a transient disturbance component; comparing the trend reference component with a preset stage evolution threshold value, and comparing the transient disturbance component with a preset tolerance boundary; according to a comparison result, mapping a current machining state to a corresponding processing branch; and storing machining control parameters and reference zero point adjustment records in isolation according to tool wear stages, so that after a tool replacement event is identified, machining control parameters and reference zero points matched with a tool state after replacement are called based on the isolated parameters and taken as initial machining parameters. The application effectively solves the technical problem that a slow drift of sensing characteristics caused by tool wear and a characteristic fluctuation caused by control parameter mismatching are mixed.

Description

Technical Field

[0001] This application relates to the field of precision metal processing technology, and more specifically, to an adaptive control method and system for a bit processing machine tool. Background Technology

[0002] In the field of precision hardware machining, in order to maintain the machining accuracy and yield of bit processing machine tools in continuous production, existing technologies usually adopt an adaptive control strategy. This involves real-time acquisition of sensor signals such as vibration, temperature and cutting force, and continuous iterative updates of control parameters such as feed rate and depth of cut based on the quality feedback after machining.

[0003] However, in actual continuous machining scenarios, as the tool gradually transitions from a sharp state to normal wear or even mid-to-late-stage wear, the change in the physical state of the tool wear directly leads to a slow but continuous unidirectional drift in the base cutting force, cutting temperature, and vibration characteristics. Existing adaptive control logic, upon receiving these continuously changing sensor characteristics, often struggles to effectively separate tool wear as an independent causative factor. The system typically misinterprets this drift caused by tool wear as a deviation of the current control parameters from their optimal settings, triggering continuous compensation adjustments. This misjudgment causes the iterative evolution of control parameters to deviate from the true optimal value, forming a continuously drifting strategy baseline. When the tool wears to its limit and needs replacement, the last set of control parameters learned by the system to compensate for wear has already significantly deviated from the reasonable operating range of the new tool. Directly using this old set of parameters will result in excessive cutting impact or accuracy anomalies in the initial tool change phase, triggering a long relearning process, ultimately causing instability in machining accuracy and a decrease in yield in the initial tool change phase.

[0004] To address the aforementioned issues, existing technologies urgently need improvement. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this application discloses an adaptive control method and system for bit processing machine tools. The aim is to solve the technical problem that existing adaptive control systems cannot effectively separate the sensor feature drift caused by tool wear from the control parameter mismatch, resulting in unstable machining accuracy and decreased yield during the initial tool change.

[0006] The technical solution of this application is as follows: In a first aspect, this application discloses an adaptive control method for a bit-machining machine tool, comprising: Obtain process state characteristics in the processing cycle; Based on process state characteristics, a trend benchmark component characterizing the evolution trend of tool wear state and a transient disturbance component characterizing instantaneous machining deviation are extracted. The trend baseline component is compared with a preset stage evolution threshold, and the transient disturbance component is compared with a preset tolerance boundary. Based on the comparison results, the current machining state is mapped to the corresponding processing branch to adjust the machining control parameters in the machining cycle and the reference zero point used to quantify the tool wear state. The adjustment records of machining control parameters and reference zero points according to the tool wear stage are isolated and stored. After a tool change event is detected, the machining control parameters and reference zero points that match the state of the replaced tool are called up as the initial machining parameters based on the isolated stored parameters.

[0007] Optionally, process state characteristics in the processing cycle are obtained, including: Collect vibration signals, temperature signals, and cutting force signals during the machining cycle; A comprehensive feature vector is generated based on vibration signals, temperature signals, and cutting force signals, serving as the process state feature.

[0008] Optionally, the adjustment records of the reference zero point and machining control parameters are stored in isolation according to the tool wear stage, including: The tool status at different wear stages is recorded using a pre-set first experience library; The reference zero point and machining control parameters in different wear stages are recorded using a pre-set second experience library to achieve logical isolation of control experience in different wear stages.

[0009] Optionally, upon detecting a tool change event, based on isolated stored parameters, machining control parameters and a reference zero point matching the state of the changed tool are retrieved as initial machining parameters, including: After identifying a tool change event, the matching wear stage is retrieved from the first experience base based on the status of the replaced tool. The reference zero point and machining control parameters corresponding to the wear stage are retrieved from the second experience library as the initial machining parameters.

[0010] Optionally, based on process state characteristics, a trend baseline component characterizing the evolution trend of tool wear state and a transient disturbance component characterizing instantaneous machining deviation are extracted, including: The process state characteristics in a continuous processing cycle are smoothed to obtain the trend baseline component. Subtract the trend reference component obtained from the previous processing cycle from the trend reference component of the current processing cycle to obtain the transient disturbance component.

[0011] Optionally, based on the comparison results, the current machining state is mapped to the corresponding processing branch to adjust the machining control parameters and reference zero point in the machining cycle, including: When all dimensions in the trend baseline component are less than the stage evolution threshold, and at least one dimension in the transient disturbance component exceeds the tolerance boundary, the transient disturbance component is used as the deviation input, and the correction compensation amount for the feed rate or cutting parameters is calculated through the preset control algorithm, while keeping the reference zero point unchanged. When at least one dimension in the trend baseline component is greater than or equal to the stage evolution threshold, and all dimensions in the transient disturbance component are within the tolerance boundary, the current processing control parameters are kept unchanged, and the current trend baseline component is confirmed as the new reference zero point. At the same time, the tolerance boundary is adjusted by shifting according to the offset of the reference zero point. When all dimensions in the trend baseline component are less than the stage evolution threshold, and all dimensions in the transient disturbance component are within the tolerance boundary, the current processing control parameters, reference zero point and tolerance boundary remain unchanged, and the process state characteristics of the next processing cycle are continuously collected. When at least one dimension of the trend baseline component is greater than or equal to the stage evolution threshold, and at least one dimension of the transient disturbance component exceeds the tolerance boundary, the tool is determined to be severely worn, a tool failure alarm signal is output, the machining cycle is paused, and the machining control parameters and reference zero point in the machining cycle are simultaneously saved.

[0012] Optionally, based on the comparison results, the current machining state is mapped to the corresponding processing branch to adjust the machining control parameters and reference zero point in the machining cycle, and the process further includes: When the trend reference component exhibits a continuous drift trend and the transient disturbance component exceeds the tolerance boundary, the current trend reference component is identified as the new reference zero point. The tolerance boundary is adjusted by shifting according to the offset of the reference zero point. Based on the adjusted tolerance boundary, the degree of exceedance of the transient disturbance component is re-evaluated and the machining control parameters are corrected.

[0013] Optionally, the method further includes: When the amplitude of at least one dimension of the transient disturbance component exceeds a preset multiple of the tolerance boundary in a single processing cycle, and falls back to within the tolerance boundary in a subsequent preset number of processing cycles, the current processing state is determined to be an occasional transient shock. When the current processing state is determined to be an occasional transient impact, the process state characteristics of the current processing cycle are removed from the calculation of the trend baseline component.

[0014] Optionally, the method further includes: The cache contains a preset number of processing cycle process state characteristics; When the amplitude of at least one dimension of the transient disturbance component in the current processing cycle exceeds a preset multiple of the tolerance boundary and does not fall back within a preset time, calculate the mean and variance of the buffered process state characteristics. If the values ​​of all dimensions of the variance are lower than the preset steady-state judgment threshold, and the value of at least one dimension of the mean is higher than the trend baseline component, then the current processing state is determined to be a step state caused by tool chipping. When the current machining state is determined to be a step state caused by tool breakage, the current trend baseline component is forcibly overwritten using the mean.

[0015] Secondly, this application also discloses an adaptive control system for a bit-machining machine tool, comprising: The feature acquisition module is used to acquire process state features in the processing cycle; The feature decomposition module is used to extract the trend benchmark component that characterizes the evolution trend of tool wear state and the transient disturbance component that characterizes the instantaneous machining deviation based on process state characteristics. The state comparison module is used to compare the trend baseline component with the preset stage evolution threshold and the transient disturbance component with the preset tolerance boundary. The parameter adjustment module is used to map the current machining state to the corresponding processing branch based on the comparison results, so as to adjust the machining control parameters in the machining cycle and the reference zero point for quantifying the tool wear state. The storage and recall module is used to isolate and store the adjustment records of machining control parameters and reference zero points according to the tool wear stage. After a tool change event is detected, the module can recall the machining control parameters and reference zero points that match the state of the replaced tool based on the isolated stored parameters, and use them as the initial machining parameters.

[0016] In summary, the adaptive control method and system for bit machining machine tools provided in this application effectively decouples the two influencing factors of tool wear and parameter mismatch by decomposing the process state characteristics during machining into a trend baseline component representing the evolution of tool wear and a transient disturbance component representing instantaneous machining deviations. The system can dynamically update the baseline of feature drift caused by tool wear based on different state mapping branches, rather than blindly compensating for parameters, thus avoiding erroneous drift in the control strategy. Simultaneously, through the isolated storage of a dual-layer experience base and a parameter reset mechanism after tool change, the system ensures the reasonable continuity of control parameters throughout the tool's entire lifecycle, completely resolving the accuracy instability problem caused by residual compensation parameters from the old tool during the initial tool change, and significantly improving the production efficiency and yield of bit machining. Attached Figure Description

[0017] Figure 1 This is a flowchart illustrating an adaptive control method for a bit-processing machine tool, as provided in an embodiment of this application.

[0018] Figure 2This is a schematic diagram of an adaptive control system for a bit processing machine tool, provided as an embodiment of this application.

[0019] Labeling Explanation: 210 Feature Acquisition Module; 220 Feature Decomposition Module; 230 State Comparison Module; 240 Parameter Adjustment Module; 250 Store and Recall Module. Detailed Implementation

[0020] The technical solutions of this application will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of this application, and not all embodiments. The components of this application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely to illustrate selected embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0021] It should be noted that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. Furthermore, in the description of this application, terms such as "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0022] In the field of precision metal parts machining, bit-making machine tools typically use the same cutting tool continuously for cutting operations when producing the same batch of workpieces. As machining continues, the tool inevitably transitions from a sharp state to normal wear or even mid-to-late-stage wear. This change in the physical state of the tool wear directly leads to a slow but continuous unidirectional drift in the cutting force base value, cutting temperature, and vibration characteristics. However, existing adaptive control logic often struggles to effectively separate tool wear as an independent causative factor when receiving these continuously changing sensor characteristics. It often misinterprets this drift in sensor characteristics caused by tool wear as a deviation of the current control parameters from their optimal settings, triggering continuous compensation adjustments, such as continuously reducing the feed rate to attempt to restore the initial stable characteristic values. This misjudgment causes the iterative evolution of control parameters to deviate from the true optimal value, forming a continuously drifting strategy baseline. When the tool wears to its limit and needs to be replaced, the last set of control parameters learned to compensate for the wear has deviated significantly from the reasonable working range of the new tool. Directly using this set of old tool parameters will cause excessive cutting impact or abnormal accuracy in the early stage of tool replacement, triggering a long relearning process.

[0023] In this regard, firstly, referring to Figure 1 This application provides an adaptive control method for a bit-machining machine tool, the method comprising: S1. Obtain the process state characteristics in the processing cycle; S2. Based on process state characteristics, extract the trend benchmark component that characterizes the evolution trend of tool wear state, and the transient disturbance component that characterizes the instantaneous machining deviation. S3. Compare the trend baseline component with the preset stage evolution threshold, and compare the transient disturbance component with the preset tolerance boundary; S4. Based on the comparison results, map the current machining state to the corresponding processing branch to adjust the machining control parameters in the machining cycle and the reference zero point used to quantify the tool wear state. S5. The adjustment records of machining control parameters and reference zero points are isolated and stored according to the tool wear stage. After a tool change event is identified, the machining control parameters and reference zero points that match the state of the replaced tool are called as the initial machining parameters based on the isolated stored parameters.

[0024] A machining cycle refers to the complete and effective cutting time window within which a bit processing machine tool completes the cutting of a single bit workpiece, from tool entry to tool exit. This time window is preferably limited to the interval during which material is actually removed between the tool and the workpiece, excluding non-cutting actions such as idle travel, shifting pauses, and fixture opening and closing from the statistical scope. This definition aims to ensure that each extracted process state characteristic corresponds to the actual cutting physical process, avoiding interference from non-cutting phases such as motor idle fluctuations, coolant jet transients, or mechanism return vibrations. For example, in the forming process of a bit workpiece, the start of the machining cycle is recorded when the tool contacts the blank and begins cutting, and the end of the machining cycle is recorded when the tool completely exits the workpiece and the cutting load returns to a stable idle state. The vibration, temperature, and cutting force information collected within this interval constitutes the machining cycle data for that workpiece.

[0025] Process state characteristics refer to the comprehensive data representation reflecting the physical response of machine tools and cutting tools within a single machining cycle. This comprehensive data representation can be a single sensor value or a feature set formed after multiple sensor values ​​have been synchronized, cleaned, normalized, and combined. Its essence is not the original signal itself, but rather an effective information carrier capable of stably characterizing the current cutting state. For example, changes in the average load in the cutting force signal can reflect the degree of tool dulling; a sudden increase in energy in a specific frequency band in the vibration signal can reflect instantaneous impact or chatter tendency; and a continuous rise in temperature can reflect increased friction and enhanced heat accumulation. The result of unifying and organizing this information can then be used as the process state characteristics input for subsequent decomposition and judgment processes.

[0026] The trend baseline component represents the long-term evolution trend of tool wear in the data space, belonging to the slow-varying baseline. Its core meaning lies in extracting the directional, cumulative, and stage-specific changes from the process state characteristics of multiple consecutive machining cycles to describe the natural evolution trajectory of the tool from new to worn. This component does not emphasize occasional anomalies in a single machining cycle, but rather the stable offsets that remain after multiple machining cycles. For example, if the average cutting force gradually increases, the steady-state temperature gradually rises, and the vibration noise gradually increases over several consecutive machining cycles, these continuously accumulating changes can be attributed to the trend baseline component.

[0027] Transient disturbance components represent immediate deviations in the physical state caused by improper feed rate settings, sudden changes in workpiece allowance, etc., during the current machining cycle; they are characterized by rapid fluctuations. These components emphasize suddenness, short duration, and recoverability. They typically do not accumulate continuously in the same direction across multiple machining cycles, but rather manifest as abnormal deviations in a single cycle or a few cycles. For example, an excessively large local allowance in a workpiece's blank can cause a temporary increase in cutting force and vibration during that machining cycle. However, this abnormality may disappear spontaneously after the workpiece's allowance returns to normal in subsequent cycles; such changes are examples of transient disturbance components.

[0028] The reference zero point is a fundamental characteristic value used to quantify the current stage of tool wear, serving as a baseline for judging whether abnormal fluctuations occur in subsequent machining processes. Its function is not to correspond fixedly to the initial state of a new tool, but rather to be updated periodically as the tool wear stage progresses, always using the normal state of the current tool stage as the comparison benchmark, rather than mistakenly using the state of a new tool as the sole standard. For example, after the tool enters the middle stage of wear, the normal levels of cutting force and temperature have generally increased compared to the new tool stage. If the characteristic values ​​of the new tool stage are still used as the judgment benchmark, it will continuously misjudge as parameter mismatch. By updating the reference zero point, this normal increase caused by wear can be absorbed into the baseline, and only deviations exceeding the normal range of the current stage are identified as abnormal.

[0029] In one possible example, acquiring process state characteristics within a machining cycle can be achieved by reading the real-time current feedback value of the machine tool spindle drive. Specifically, the spindle motor current data can be continuously read at a high-frequency cycle via the machine tool's internal communication bus. This high-frequency cycle should be shorter than the characteristic time of significant changes in cutting load within a single machining cycle to ensure that load changes during the entry, stable cutting, and exit phases can be captured. To enable this acquisition process to directly serve subsequent state analysis, it is preferable to receive a cycle start marker from the CNC system or stroke trigger signal before each machining cycle begins. The current sequence should be buffered after the tool enters the workpiece cutting zone and buffered after the tool exits the workpiece and returns to no-load operation, thus obtaining a current time sequence corresponding one-to-one with a single machining cycle. Subsequently, using a pre-calibrated current-torque mapping model, the current data is converted into cutting load characteristics as process state characteristics.

[0030] To more clearly and completely describe this implementation method, the pre-calibrated current-torque mapping model can be obtained as follows: Under the conditions that the machine tool is in a healthy state, the transmission mechanism is properly lubricated, the tool is properly installed, and the workpiece material batches are stable, several representative machining conditions are selected, and the feedback current of the spindle drive and the corresponding actual load changes are recorded. The actual load changes can be obtained through the machine tool's existing torque monitoring interface, an external standard force measuring device, or through comparative calibration under known conditions such as no-load, light-load, and normal cutting. A correspondence is established between the current values ​​under different conditions and the corresponding load states, forming a mapping table or segmented mapping rules. For example, the no-load current baseline is recorded during the no-load rotation stage, the load current rise is recorded during the stable cutting stage, and the current change range is recorded under different material removal intensities. These correspondences are then solidified into a callable mapping model. During actual operation, the currently collected current value is first subtracted from the no-load baseline under the same speed condition, and then converted into cutting load characteristics according to the mapping model. The reason for this approach is that the spindle current contains both the base current required for the motor to maintain rotation and the additional load current caused by cutting. If the no-load baseline is not eliminated first, changes in spindle bearing friction, lubrication status, or transmission resistance can easily be mistaken for changes in cutting load.

[0031] To isolate and store the adjustment records of machining control parameters and reference zero points according to tool wear stages, the parameter adjustment records of all stages can be written to a unified global log file in chronological order. In practice, after each adjustment of machining control parameters or reference zero points, a data record containing a timestamp, the current tool wear stage identifier, the parameters before adjustment, and the parameters after adjustment is generated and appended to the end of the global log file. Upon detecting a tool change event, the log file is traversed, and matching initial machining parameters are selected based on the timestamp and stage identifier.

[0032] To make this implementation more practical, in addition to timestamps, current tool wear stage identifiers, parameters before and after adjustment, data records can further include tool identifiers, workpiece batch identifiers, current reference zero-point version numbers, triggering adjustment reason identifiers, and a summary of the process state characteristics corresponding to this adjustment. This is because relying solely on timestamps and stage identifiers for retrieval can easily lead to mixed search results, especially in cases of mid-process shutdowns, cross-batch production, manual trial cutting, or repeated clamping. Adding tool identifiers and reference zero-point version numbers allows for clearer differentiation of adjustment records under different tool entities or different baseline versions within the same stage. Tool identifiers can originate from manual input in the tool change operation interface, the tool holder identification module, tool life management numbers, or be automatically generated as a new tool session number after a tool change event is detected. The triggering adjustment reason identifier distinguishes whether the record's reference zero-point update is due to a trend baseline component crossing a stage evolution threshold, or a machining control parameter correction due to a transient disturbance component exceeding the tolerance boundary.

[0033] Furthermore, after identifying a tool change event, instead of simply tracing back from the end of the global log file, the system first determines the search criteria that match the state of the replaced tool. These criteria include at least an initial wear stage identifier for the new tool and process context information consistent with or similar to the current machining task, such as workpiece type, material category, tool model, and machining program version. Subsequently, historical records meeting these criteria are filtered from the global log file, with priority given to a set of records under similar process conditions, in the initial stage of a new tool, and with a longer stable running time, as candidate initial machining parameters. If multiple sets of candidate records exist, their corresponding process state feature summaries can be further compared, and the set with the smallest deviation from the current start-up trial cut result is selected as the call result. If no completely matching record is found, the preset conservative initial parameters are called, and the reference zero point of the tool is re-established in several subsequent machining cycles. In this way, even using the relatively direct storage format of a global log file, parameter backtracking and baseline restoration after a tool change can still be achieved. However, its search efficiency and stage isolation effect are relatively weaker than the dual-experience-base scheme, making it more suitable as a feasible basic implementation method or an alternative implementation.

[0034] Through the aforementioned overall technical solution, this application effectively solves the technical problem of mutual confusion between slow drift of sensing characteristics caused by tool wear and characteristic fluctuations caused by control parameter mismatch. Its core logic lies in decomposing process state characteristics into trend baseline components and transient disturbance components, thus achieving complete decoupling of the causative factors at the logical level. When characteristics change, instead of blindly attributing them all to parameter mismatch, it accurately locates whether the current deviation stems from natural tool wear or short-term machining disturbances by comparing it with stage evolution thresholds and tolerance boundaries. For slow drift caused by tool wear, the reference zero point is updated to proactively adapt to the new tool state, blocking the continuous error compensation chain; for rapid disturbances caused by parameter mismatch, precise parameter correction is performed. More importantly, through an isolated storage mechanism, the initial machining parameters matching the new tool state can be directly recalled after tool change, completely eliminating the problem of machining accuracy instability and yield reduction caused by using incorrect parameters from the old tool in the early stages of tool change, achieving efficient and stable adaptive control of the bit processing machine tool throughout its entire lifecycle.

[0035] Optionally, process state characteristics in the processing cycle are obtained, including: Collect vibration signals, temperature signals, and cutting force signals during the machining cycle; A comprehensive feature vector is generated based on vibration signals, temperature signals, and cutting force signals, serving as the process state feature.

[0036] Specifically, a broadband piezoelectric triaxial accelerometer can be installed at a rigid node near the front bearing on the machine tool spindle housing to capture the mechanical vibration signals generated when the tool cuts the workpiece in real time with a high-frequency sampling rate. Simultaneously, an infrared thermal imaging sensor is deployed near the coolant nozzle to continuously collect temperature signals from the cutting zone. Furthermore, a triaxial piezoelectric force gauge is integrated at the bottom of the worktable to directly acquire cutting force signals. The placement of these sensors is not arbitrary but rather chosen to be as close as possible to the cutting energy transmission path while avoiding installation positions that could compromise the rigidity of the machine tool structure and ease of maintenance. The rigid node on the spindle housing near the front bearing typically transmits cutting vibrations more easily and experiences less secondary resonance from distant structural components, making it suitable for acquiring vibration signals. The area near the coolant nozzle allows for direct observation of thermal radiation changes in the cutting zone without exposing the temperature sensor directly to the tool contact area. Placing the force gauge at the bottom of the worktable or along the force transmission path of the fixture facilitates the acquisition of overall force changes related to the cutting load.

[0037] Within each effective cutting time window, the acquired vibration, temperature, and cutting force signals are strictly timestamped and then cleaned using methods such as mean removal and noise reduction. Strict timestamping means mapping data from different acquisition channels to the same time reference within the same machining cycle, ensuring that vibration, temperature, and cutting force changes at any given moment correspond to each other. If different sensors have different sampling cycles, synchronization is first achieved using a unified clock or acquisition card time reference, and then high-sampling-rate signals are aggregated by time slices to an analysis granularity consistent with low-sampling-rate signals. Mean removal primarily eliminates sensor static bias, while noise reduction suppresses environmental noise, coolant splash interference, electromagnetic crosstalk, and irrelevant fluctuations caused by non-cutting actions. For vibration signals, frequency bands related to cutting contact, friction, and impact are preferentially retained; for temperature signals, background heat sources at the field of view edges are eliminated, retaining only the effective temperature rise area near the cutting zone; for cutting force signals, constant biases caused by clamping preload and platform static load are removed.

[0038] Subsequently, the mean and variance of the cutting force, the average steady-state temperature of the temperature field, and the frequency band energy of the vibration signal are extracted, and these multi-source time-frequency features are combined into a high-dimensional comprehensive feature vector. To ensure that this comprehensive feature vector can be stably used for subsequent decomposition, it is preferable to first unify the dimensions and normalize the amplitude of each type of feature, and then assemble them into a data object with a fixed structure according to a predetermined order. For example, mechanically relevant features can be included first, followed by thermally relevant features, and finally vibration-related features, with a source identifier and acquisition confidence identifier added to each dimension. If a sensor experiences occlusion, saturation, detachment, or communication abnormalities in the current machining cycle, a missing marker can be added to the corresponding feature location, and the confidence level of the overall features of that machining cycle can be reduced, rather than directly using outliers in subsequent judgments. This approach avoids the single sensor failure skewing the entire comprehensive feature vector.

[0039] Furthermore, the specific processing method for generating a comprehensive feature vector based on vibration signals, temperature signals, and cutting force signals can be implemented as follows: First, identify stable cutting sections within a single machining cycle, and then extract summary features representing the state of that section from the three signal sources. For the cutting force signal, extract the average load level, fluctuation amplitude, and whether there is a sudden impact peak; for the temperature signal, extract the average temperature rise level of the cutting zone, the duration of the temperature rise, and whether there are local abnormal hot spots in the heat distribution; for the vibration signal, extract the overall vibration intensity, specific frequency band energy changes, and whether periodic impact features appear. Then, write these summary features into the comprehensive feature vector in a unified order, and bind and store this vector with the current machining cycle number. Next, observe the change trajectory of this comprehensive feature vector over multiple consecutive machining cycles: if the average cutting force load, average temperature rise, and vibration noise floor all show a continuous increase in multiple cycles, then this continuous increase is assigned to the trend baseline component; if only one cycle shows an impact peak, local hot spot, or a sudden increase in vibration frequency band energy, and this does not continue into the preceding and following cycles, then this part is assigned to the transient disturbance component. By constructing a comprehensive feature vector first and then performing time-scale decomposition, the physical meaning of the trend baseline component and the transient disturbance component becomes clearer, and the separation results are more stable.

[0040] In some preferred embodiments, the adjustment records of the reference zero point and machining control parameters are stored in isolation according to the tool wear stage, including: The tool status at different wear stages is recorded using a pre-set first experience library; The reference zero point and machining control parameters in different wear stages are recorded using a pre-set second experience library to achieve logical isolation of control experience in different wear stages.

[0041] For specific examples, refer to the following instance. The first experience base serves as a tool stage state trajectory database, using the cumulative machining time or number of machined parts as the primary key. It periodically and persistently stores the currently determined tool wear stage identifier and the corresponding macro-level tool status. The cumulative machining time or number of machined parts does not need to be used simultaneously; in actual deployment, the choice can be made based on the data types that the machine tool control system can reliably acquire. If the machining cycle time is stable and the difference in single-piece cutting time is small, the cumulative number of machined parts is sufficient to characterize the tool life progression. If the cutting time varies significantly for different bit specifications, the cumulative effective cutting time is more conducive to accurately reflecting the wear process. The macro-level tool status can include discrete stage identifiers such as the initial stage of a new tool, the early break-in stage, the stable wear stage, and the mid-to-late wear stage. It can also include status descriptions corresponding to these stages, such as suggesting increased monitoring frequency closer to the tool change window for normal production. The role of the first experience base is to record the main trajectory of the tool lifecycle, clearly indicating which wear stage the tool is in at a given moment.

[0042] The second experience base serves as a historical database for fine-tuning parameters at the current stage. Indexed by the characteristics of the current transient disturbance components, it records updated reference zero points and corrected machining control parameters for specific tool wear stages, designed to address rapidly changing disturbances. To avoid confusion, the recorded reference zero points do not imply that the second experience base defines the tool stage itself. Rather, it stores confirmed effective reference zero point versions and their associated control parameters within a specific wear stage, allowing for rapid reuse of existing adjustments when encountering similar disturbances within the same stage. Machining control parameters may include feed rate, spindle speed, depth of cut correction, cooling strategy selection, or other parameters compatible with the current machine tool control interface. If certain parameters are not supported for online adjustment on a specific machine tool, only those that can be adjusted online are recorded. Using transient disturbance component characteristics as an index means organizing information such as disturbance type summary, disturbance strength level, occurrence station or process segment, and corresponding workpiece batch context into searchable tags, without requiring a complex mathematical index structure. With this setting, when a similar transient disturbance is encountered again at the same wear stage, a parameter correction scheme that has been proven effective in the past can be retrieved directly, without having to search from scratch.

[0043] To more clearly and completely illustrate the formation of the first and second experience bases, the following example can be used: In the initial deployment phase, several cutting tools of the same model are selected and continuously produced under the same or similar bit processing tasks. The process state characteristics, trend baseline components, transient disturbance components, reference zero-point changes, and actual processing quality results are recorded synchronously for each processing cycle. As the cutting tools gradually progress from new tools to different wear stages, those stages that stably correspond to a certain wear stage and are consistent with the processing quality results are written into the first experience base, forming an initial stage trajectory template. Simultaneously, within each wear stage, the parameter correction actions taken for different transient disturbances and their subsequent effects are written into the second experience base, and the results of whether stability is restored after several processing cycles are retained. After multiple rounds of production accumulation, parameter correction records that can repeatedly achieve stable effects under similar disturbances are retained as valid experience in the second experience base; while records with unstable effects, accompanied by quality fluctuations, or only valid under individual abnormal conditions are marked as low priority or not included in automatic recall.

[0044] In some preferred embodiments, after a tool change event is detected, based on isolated stored parameters, machining control parameters and a reference zero point that match the state of the changed tool are retrieved as initial machining parameters, including: After identifying a tool change event, the matching wear stage is retrieved from the first experience base based on the status of the replaced tool. The reference zero point and machining control parameters corresponding to the wear stage are retrieved from the second experience library as the initial machining parameters.

[0045] To ensure the stable implementation of this call process, tool change events can be identified using at least one reliable information acquisition method. The first method is to monitor the tool change command code in the CNC system. When the CNC system executes the automatic tool change program and outputs the corresponding tool change completion signal, this signal is recorded as a tool change event. The second method is to read the RFID tag on the tool holder or the identification information written by the tool presetter. By reading fields such as tool number, tool type, tool diameter, cutting edge length, regrinding count, and preset lifespan level, it can be confirmed that the currently used tool has changed. The third method is for the tool magazine management module to send a confirmation signal to the upper-level control module after the robot arm completes the gripping, insertion, and locking of the tool. In actual deployment, cross-verification of CNC system signals and RFID information can be prioritized to reduce the probability of false identification. The status of the replaced tool can include at least one or more of the following: new tool, regrinded tool, replacement tool of the same model, tool from the same batch, tool from a different batch, estimated lifespan level, cutting edge compensation status, and cumulative historical machining value of the tool. The retrieval fields in the first experience database can be recorded according to the format of tool type—tool status—wear stage identifier. The records in the second experience base can be established in the format of wear stage identifier - reference zero point - machining control parameters - applicable working condition description, so that the corresponding parameters can be directly called after the wear stage is identified, without having to recalculate from a large amount of raw data.

[0046] In the specific implementation process, the occurrence of a tool change event is first accurately identified by receiving a tool change command code or scanning an RFID tag on the tool holder, and the status information of the replaced tool is obtained. Then, using this tool status information as a search condition, the system accesses the first experience database, which serves as a macroscopic state trajectory library, to match and determine the current wear stage identifier of the tool. After identifying the wear stage, the system carries this stage identifier into the second experience database, which serves as a historical database of microscopic parameters. It then precisely retrieves the historically verified optimal reference zero point and associated machining control parameters for that specific wear stage and sends them to the machine tool's underlying control system.

[0047] To describe this implementation process more clearly and completely, it can be executed in the following order: First, the controller reads the tool identification information within a preset time window after receiving the tool change completion signal, avoiding premature sampling before the robot arm has completely exited; second, the read tool identification information is validated, such as checking whether the tool number exists in the tool master file, whether the tool type is consistent with the current machining program, and whether the regrinding mark is valid; then, the valid tool status information is input into the retrieval interface of the first experience base to retrieve the unique wear stage identifier. If there are multiple candidate results, they are filtered in the order of tool type priority, most recent status record priority, and historical machining consistency priority; after obtaining the wear stage identifier, the second experience base is accessed to read the reference zero point and machining control parameters that have been encapsulated for this stage; finally, these parameters are written into the CNC system variable area, tool compensation area, or the parameter cache area of ​​the upper controller, so that these parameters are effective before the start of the next machining cycle. To prevent the misuse of invalid data in the experience base, each parameter record in the second experience base can also be accompanied by a validity mark, the last update time, the applicable material range, the applicable process number, and the data source batch. Only records that meet the conditions of the current processing task are allowed to be issued.

[0048] For example, after the automatic tool changer completes the tool change, it confirms by reading the tool holder chip that a regrinded, semi-new tool has been installed. First, it accesses the first experience database, using the regrinded tool status tag to retrieve the corresponding wear stage identifier as the early break-in stage. Then, using the early break-in stage identifier as an index, it queries the second experience database, retrieving the most stable initial cutting force baseline from the historical records for that stage as the reference zero point, and retrieving the corresponding initial feed rate and spindle speed as machining control parameters, directly applying them to the upcoming machining cycle. To make this example feasible, the data sources can be further explained: the tool holder chip can be pre-written with the tool number, whether the tool has been regrinded, the number of regrinding cycles, and the most recent pre-adjustment information; the first experience base's stage division for regrinding tools can be derived from tracking records of multiple tools of the same type during early batch processing. These records include changes in cutting force, temperature, vibration, and surface quality for each tool during the initial few machining cycles after regrinding. After manual review, states that meet the requirements for short-term re-running are uniformly classified into the early run-in stage; the second experience base's reference zero point and machining control parameters corresponding to the early run-in stage can be derived from data segments showing stable performance in normal machining from multiple batches of regrinded tools in history. This is formed after filtering out abnormal samples such as tool chipping, abnormal workpiece clamping, and excessive material fluctuations. Thus, when the current tool is identified as a regrinding tool, it doesn't simply use entirely new tool parameters, but rather uses parameters more suitable for the initial state of the regrinding tool, thereby reducing the impact and misadjustment after tool change.

[0049] Optionally, based on process state characteristics, a trend baseline component characterizing the evolution trend of tool wear state and a transient disturbance component characterizing instantaneous machining deviation are extracted, including: The process state characteristics in a continuous processing cycle are smoothed to obtain the trend baseline component. Subtract the trend reference component obtained from the previous processing cycle from the trend reference component of the current processing cycle to obtain the transient disturbance component.

[0050] Specifically, smoothing does not directly process a single original sampling point, but rather processes the feature sequence extracted from multiple consecutive machining cycles, thus better reflecting the actual characteristic of tool wear as a gradual process across cycles. Smoothing algorithms such as exponentially weighted moving averages or sliding window filtering can be used to perform low-pass filtering on the process state features of multiple consecutive machining cycles. The purpose of this processing is to filter out high-frequency random noise and retain low-frequency long-term evolution trends, thereby obtaining a stable trend baseline component characterizing the tool wear state.

[0051] Furthermore, the difference between the trend baseline component of the current processing cycle and the trend baseline component updated in the previous processing cycle is calculated. Mathematically, this difference represents the rate of change or increment of the trend baseline component, and its physical meaning lies in quantifying the degree to which the latest added process state feature pulls on the overall smoothed baseline. When there is a significant immediate processing deviation, the current feature will force the smoothed baseline component to produce a sudden increment; therefore, this difference can sensitively and indirectly characterize the transient disturbance component in the current processing cycle.

[0052] To illustrate the implementation more clearly and completely, the process state features extracted from each machining cycle can be written into the cycle feature queue after each machining cycle ends. If an exponentially weighted moving average is used, the most recently updated value is given a higher weight, and the earlier historical value is given a lower weight, so that the trend baseline component can remain smooth and gradually follow the changes in tool wear. If a sliding window filter is used, the smoothing result is calculated in a fixed number of recent machining cycles, and the window length can be preset according to the tool wear rate and machining cycle time.

[0053] In one specific embodiment, an exponentially weighted moving average algorithm is used to smooth the cutting force characteristics of multiple consecutive machining cycles. Assume the cutting force trend baseline component updated in the previous machining cycle is 100N, and the actual cutting force characteristic collected in the current machining cycle reaches 150N due to a sudden increase in the blank allowance. After exponentially weighted smoothing, the cutting force trend baseline component of the current machining cycle is increased to 110N. Then, the trend baseline component (100N) updated in the previous machining cycle is subtracted from the trend baseline component (110N) of the current machining cycle, resulting in a difference of 10N. This 10N difference is extracted as the transient disturbance component, used to characterize the immediate machining deviation caused by the sudden increase in the blank allowance in the current machining cycle. In this example, the initial cutting force characteristic instantaneously increased from around 100N to 150N, indicating a significant deviation in this machining cycle. However, due to the buffering effect of the smoothing algorithm, the trend baseline component is not completely pulled to 150N by a single sudden anomaly, but only rises from 100N to 110N. This indicates that the event is still considered a short-term anomaly rather than an immediate jump in overall tool wear. The subsequent 10N difference reflects the additional push that the new data has on the smoothing baseline. If the cutting force falls back in several subsequent machining cycles, the difference will also return to a smaller level, indicating that the anomaly is an instantaneous disturbance. If the cutting force remains high for several subsequent machining cycles, the trend baseline component will continue to gradually rise, and this will be combined with subsequent threshold comparisons to determine whether a new wear stage has begun. Therefore, this processing method retains the ability to perceive immediate anomalies while avoiding overreaction to single anomalies.

[0054] In practical applications, based on the comparison results, the current machining state is mapped to the corresponding processing branch to adjust the machining control parameters and reference zero point in the machining cycle, including: When all dimensions in the trend baseline component are less than the stage evolution threshold, and at least one dimension in the transient disturbance component exceeds the tolerance boundary, the transient disturbance component is used as the deviation input, and the correction compensation amount for the feed rate or cutting parameters is calculated through the preset control algorithm, while keeping the reference zero point unchanged. When at least one dimension in the trend baseline component is greater than or equal to the stage evolution threshold, and all dimensions in the transient disturbance component are within the tolerance boundary, the current processing control parameters are kept unchanged, and the current trend baseline component is confirmed as the new reference zero point. At the same time, the tolerance boundary is adjusted by shifting according to the offset of the reference zero point. When all dimensions in the trend baseline component are less than the stage evolution threshold, and all dimensions in the transient disturbance component are within the tolerance boundary, the current processing control parameters, reference zero point and tolerance boundary remain unchanged, and the process state characteristics of the next processing cycle are continuously collected. When at least one dimension of the trend baseline component is greater than or equal to the stage evolution threshold, and at least one dimension of the transient disturbance component exceeds the tolerance boundary, the tool is determined to be severely worn, a tool failure alarm signal is output, the machining cycle is paused, and the machining control parameters and reference zero point in the machining cycle are simultaneously saved.

[0055] Specifically, the preset control algorithm can be established through preliminary trial cuts. Taking feed rate compensation as an example, the correspondence between cutting force deviation and machining quality changes can be collected under different materials, different tool wear stages, and different disturbance intensities. Then, a control table for the disturbance deviation range and feed rate correction strategy can be formed. Alternatively, a verified and stable proportional-integral-derivative (PID) control algorithm can be used to convert the deviation into a feed rate correction amount. For example, the following types of mapping records can be pre-established: slight adjustment of feed rate for minor overshoot, moderate reduction of feed rate for moderate overshoot, and simultaneous reduction of feed rate and increase of coolant supply level for multiple consecutive overshoots. For updating the reference zero point, the current trend benchmark component can be written into the baseline storage area of ​​the corresponding feature dimension, and the tolerance boundary set around the old reference zero point can be shifted as a whole to a position centered on the new reference zero point to ensure that the normal range on which subsequent judgments are based changes synchronously with the natural wear of the tool. The machining control parameters and reference zero point in the sealed machining cycle can be understood as saving the control parameters, reference zero point, trend reference component, transient disturbance component, alarm type, tool number and workpiece batch at the current moment to the isolated storage area. This prevents these extreme abnormal data from directly entering the normal sample set in the experience library, but they can be used for subsequent failure analysis and maintenance traceability.

[0056] In practical implementation, the first branch corresponds to the parameter mismatch condition, indicating that the tool has not entered the next wear stage, but there is an immediate machining anomaly. In this case, only the control parameters are corrected without updating the baseline. The second branch corresponds to the normal wear condition, indicating that the tool's natural wear accumulation has reached the stage transition point and the current machining is stable. In this case, only the reference zero point and translation tolerance boundary are updated without modifying the control parameters to avoid erroneous compensation. The third branch corresponds to the stable cutting condition, indicating that everything is normal and should be maintained as is. The fourth branch corresponds to the tool failure condition, indicating that the tool is not only severely worn but has also caused violent immediate fluctuations. In this case, the machine must be stopped immediately, an alarm must be triggered, and abnormal data must be sealed. For ease of implementation, each feature dimension can be judged first, and then the overall status label can be obtained by summarizing. For example, in a scenario where both cutting force and temperature are monitored simultaneously, as long as any trend baseline component of cutting force or temperature reaches the stage evolution threshold, the condition that at least one dimension is greater than or equal to the stage evolution threshold is satisfied; as long as the transient disturbance component exceeds the tolerance boundary in any dimension, the condition that at least one dimension exceeds the tolerance boundary is satisfied.

[0057] In one specific embodiment, during a batch continuous machining task, cutting force and temperature characteristics are monitored in real time. During the 500th machining cycle, it is found that the cutting force trend reference component has not reached the stage evolution threshold, but due to the high local hardness of the workpiece, the transient disturbance component of the cutting force exceeds the tolerance boundary. The first processing branch is initiated, maintaining the reference zero point unchanged, and using a proportional-integral-derivative (PID) control algorithm to calculate the feed rate reduction based on the disturbance deviation, thus lowering the feed rate to smoothly navigate the high-hardness region. During the 2000th machining cycle, with normal tool wear, the temperature trend reference component slowly rises and reaches the stage evolution threshold, while the transient disturbance component remains within the tolerance boundary. The second processing branch is initiated, determining that the tool has entered the next wear stage. Therefore, the current feed rate is maintained, the current temperature trend reference component is confirmed as the new reference zero point, and the overall temperature tolerance boundary is shifted upwards to accommodate the normal temperature rise after wear. During the 3500th machining cycle, the tool flank wear deteriorated sharply, causing the cutting force trend reference component to significantly exceed the threshold. Simultaneously, severe vibration caused the transient disturbance component to also severely exceed the tolerance boundary. Entering the fourth processing branch, the system immediately determined severe tool wear, outputting a tool failure alarm signal to the machine tool control panel, triggering an emergency spindle stop to pause the machining cycle, and saving the current machining control parameters and reference zero point to prevent these extreme anomalies from contaminating the experience database.

[0058] To make this example more complete, the connection between each action can be further explained: In the first processing branch, the execution result after the feed rate is reduced can be used to re-collect process state characteristics in the next machining cycle to verify the effectiveness of the compensation; if the anomaly disappears, it reverts to the stable cutting or normal wear branch. Once the new reference zero point and new tolerance boundary are confirmed in the second processing branch, they replace the old baseline as the basis for judgment in subsequent cycles. The archived data in the fourth processing branch is preferably written to the anomaly event archive and marked as not suitable for direct input into the training database to avoid mistaking data under failure states as normal experience samples.

[0059] In a preferred embodiment of this application, based on the comparison result, the current processing state is mapped to the corresponding processing branch to adjust the processing control parameters and reference zero point in the processing cycle, and the method further includes: When the trend reference component exhibits a continuous drift trend and the transient disturbance component exceeds the tolerance boundary, the current trend reference component is identified as the new reference zero point. The tolerance boundary is adjusted by shifting according to the offset of the reference zero point. Based on the adjusted tolerance boundary, the degree of exceedance of the transient disturbance component is re-evaluated and the machining control parameters are corrected.

[0060] The rules for determining whether a trend reference component exhibits a continuous drift trend can be preset and stored. For example, it can be checked whether the trend reference component continuously changes in the same direction over multiple consecutive machining cycles, and a comprehensive judgment can be made based on factors such as whether the cumulative change reaches a preset proportion, whether the change is continuous, and whether there is a significant reverse fall. These preset rules can be derived from offline analysis of long-term trial cutting data: typical continuous drift stages are selected from the full-life machining data of multiple batches of tools, their common characteristics in cutting force, temperature, etc., are statistically analyzed, and then a rule table for determining drift is formed. The process of re-evaluating transient disturbance components can be performed as follows: first, the current trend reference component is written to the reference zero-point storage area; then, the tolerance boundary is moved as a whole to a position around the new reference zero-point; next, using the new tolerance boundary as the judgment benchmark, the current transient disturbance component is checked to see if it is slightly, moderately, or severely exceeding the limit, and the compensation intensity of the control algorithm output is reselected accordingly. Therefore, instead of directly compensating for the exceeding results under the old baseline, the actual disturbance degree is re-measured after the baseline is corrected, making the compensation action closer to the actual working conditions.

[0061] In practice, the historical trajectory of the trend reference component is monitored in real time. When it is found that the trend reference component of multiple consecutive machining cycles monotonically increases in the same direction and the cumulative change reaches a preset proportion, it is determined that it exhibits a continuous drift trend. In this case, the original reference zero point has seriously lagged behind the actual physical baseline. Therefore, the currently raised trend reference component is first identified as the new reference zero point, and the tolerance boundary is simultaneously shifted. Subsequently, under the new tolerance boundary, the difference between the transient disturbance component and the new boundary is recalculated. The transient disturbance that seemed serious under the old boundary may only be a slight exceedance under the new boundary, and the correction amount of the machining control parameters is recalculated accordingly. To explain this logic more clearly, it can be understood as a two-step process: the first step is to solve the scale inaccuracy problem, that is, the original reference zero point can no longer represent the normal state of the current tool; the second step is to solve the actual deviation problem, that is, how much the current machining cycle deviates under the updated normal state. In terms of information acquisition, the existence of continuous drift can be determined by saving the trend reference component sequence of the most recent machining cycle. The saving length can be set according to the machining cycle and tool wear rate. In terms of information processing, the historical trajectory can be updated after each new cycle, and the criteria for continuous drift can be checked. This supplementary mechanism is only triggered when both the continuous drift trend and the transient disturbance component exceed the tolerance boundary, in order to avoid resetting the baseline too frequently for normal operating conditions.

[0062] In one specific embodiment, when machining a batch of high-hardness alloy bits, the tool wear rate is relatively fast. The cutting force trend reference component was monitored to exhibit a continuously increasing drift trend over 30 consecutive machining cycles, while the transient disturbance component of the cutting force in the current machining cycle exceeded the original tolerance boundary upper limit. At this point, instead of directly calculating the speed reduction compensation based on the original tolerance boundary, the currently raised cutting force trend reference component was first identified as the new reference zero point, and the tolerance boundary was shifted upwards by 50N. After completing the dynamic reset of the baseline and boundary, the degree of exceedance of the current transient disturbance component was recalculated based on the raised new tolerance boundary. The results showed that under the new boundary, the transient disturbance component only slightly exceeded the limit by 10N, instead of the previously estimated 60N. Therefore, only a slight downward adjustment was made to the feed rate, avoiding a significant drop in feed rate due to misjudgment. The 30 consecutive machining cycles and the upward shift of the tolerance boundary by 50N in this example can be understood as a set of process settings that have been proven effective for machining high-hardness alloy bits: the former illustrates that continuous drift is not a short-term phenomenon, but rather a sufficiently continuous wear evolution; the latter illustrates that after the reference zero point is raised, the tolerance boundary needs to be moved upward synchronously to maintain consistency between the judgment range and the current tool state. If used with other materials or other tool specifications, the corresponding preset values ​​can be adopted according to the same principle without changing the processing mechanism itself.

[0063] In a preferred embodiment of this application, the method further includes: When the amplitude of at least one dimension of the transient disturbance component exceeds a preset multiple of the tolerance boundary in a single processing cycle, and falls back to within the tolerance boundary in a subsequent preset number of processing cycles, the current processing state is determined to be an occasional transient shock. When the current processing state is determined to be an occasional transient impact, the process state characteristics of the current processing cycle are removed from the calculation of the trend baseline component.

[0064] Specifically, when an extreme exceedance of transient disturbance components is detected (i.e., exceeding a preset multiple of the tolerance boundary, such as 2 or 3 times), it is not immediately attributed to severe tool failure or continuous drift. Instead, a short observation window is initiated. The length of this observation window is the preset number of machining cycles, typically set to 3 to 5 cycles. If, within this observation window, the transient disturbance components of subsequent machining cycles quickly fall back to within the normal tolerance boundary, it indicates that the aforementioned extreme exceedance was merely an isolated physical event, i.e., an occasional transient impact. In this case, the abnormal machining cycle is located in the historical data cache, and its corresponding process state characteristics are removed from the smoothing queue used to calculate the trend baseline components, or replaced with interpolation from the preceding and following normal cycles.

[0065] In another preferred embodiment of this application, the method further includes: The cache contains a preset number of processing cycle process state characteristics; When the amplitude of at least one dimension of the transient disturbance component in the current processing cycle exceeds a preset multiple of the tolerance boundary and does not fall back within a preset time, calculate the mean and variance of the buffered process state characteristics. If the values ​​of all dimensions of the variance are lower than the preset steady-state judgment threshold, and the value of at least one dimension of the mean is higher than the trend baseline component, then the current processing state is determined to be a step state caused by tool chipping. When the current machining state is determined to be a step state caused by tool breakage, the current trend baseline component is forcibly overwritten using the mean.

[0066] In practical implementation, a fixed-length sliding window can be continuously maintained in memory to cache the process state features of the most recent preset number of machining cycles. When a transient disturbance component is detected to exceed its limit and fails to return to normal after the aforementioned short observation window (i.e., preset time), it is determined that a continuous abrupt change has occurred in the current working condition. At this time, statistical analysis is performed on the process state features within the cache window to calculate their mean and variance. The variance is used to measure whether the machining state after the abrupt change has returned to stability: if the values ​​of all dimensions of the variance are lower than the preset steady-state judgment threshold, it indicates that although the amplitude of the features after the abrupt change has increased, the fluctuation is small, and a new stable platform has been formed; the mean is used to confirm the direction of the abrupt change: if the mean is significantly higher than the original trend baseline component, combined with the stable performance of the variance, it can be accurately determined that a physical event such as tool chipping has occurred, leading to a step increase in cutting load. After confirming the step state, instead of relying on a slow smoothing algorithm to gradually approach the new platform, the calculated mean is directly used to forcibly replace (overwrite) the current trend baseline component, thereby completing the baseline update.

[0067] Based on the control logic and processing mechanism provided in the above method embodiments, in the second aspect, referring to... Figure 2 This application also proposes an adaptive control system for a bit-machining machine tool, for performing the steps in any of the above methods, including: Feature acquisition module 210 is used to acquire process state features in the processing cycle; The feature decomposition module 220 is used to extract the trend reference component that characterizes the evolution trend of tool wear state and the transient disturbance component that characterizes the instantaneous machining deviation based on the process state features. The state comparison module 230 is used to compare the trend baseline component with a preset stage evolution threshold and to compare the transient disturbance component with a preset tolerance boundary. The parameter adjustment module 240 is used to map the current machining state to the corresponding processing branch based on the comparison results, so as to adjust the machining control parameters in the machining cycle and the reference zero point for quantifying the tool wear state. The storage and recall module 250 is used to isolate and store the adjustment records of machining control parameters and reference zero points according to the tool wear stage. After a tool change event is detected, the machining control parameters and reference zero points that match the state of the replaced tool are recalled based on the isolated stored parameters as the initial machining parameters.

[0068] Through the above technical solution, this application solidifies the complex adaptive control method into a highly modular system architecture. The feature acquisition module 210 interfaces with the underlying sensors and CNC system, ensuring real-time acquisition and synchronization of multi-source data; the feature decomposition module 220 achieves precise decoupling between slow-changing trends and fast-changing disturbances; the state comparison module 230 and the parameter adjustment module 240 work together to form a multi-dimensional state mapping and decision core, ensuring the accuracy and appropriateness of control actions; the storage and retrieval module 250 completely solves the problem of blind parameter retrieval after tool change through the logical isolation of a two-layer experience base. The data flow between modules is clear, and the boundaries of responsibilities are well-defined, jointly endowing the bit processing machine tool with efficient, stable, and intelligent adaptive control capabilities throughout its entire life cycle.

[0069] The above are merely embodiments of this application and are not intended to limit the scope of protection of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. An adaptive control method for a bit-machining machine tool, characterized in that, The method includes: Obtain process state characteristics in the processing cycle; Based on the process state characteristics, a trend benchmark component characterizing the evolution trend of tool wear state and a transient disturbance component characterizing instantaneous machining deviation are extracted. The trend baseline component is compared with a preset stage evolution threshold, and the transient disturbance component is compared with a preset tolerance boundary. Based on the comparison results, the current machining state is mapped to the corresponding processing branch to adjust the machining control parameters in the machining cycle and the reference zero point used to quantify the tool wear state. The adjustment records of the machining control parameters and the reference zero point are isolated and stored according to the tool wear stage. After a tool change event is detected, the machining control parameters and reference zero point that match the state of the replaced tool are called as the initial machining parameters based on the isolated stored parameters.

2. The adaptive control method for a bit-machining machine tool according to claim 1, characterized in that, The acquisition of process state characteristics in the processing cycle includes: Collect vibration signals, temperature signals, and cutting force signals during the machining cycle; A comprehensive feature vector is generated based on the vibration signal, the temperature signal, and the cutting force signal, which serves as the process state feature.

3. The adaptive control method for a bit-machining machine tool according to claim 1, characterized in that, The method of isolating and storing the adjustment records of the reference zero point and the machining control parameters according to the tool wear stage includes: The tool status at different wear stages is recorded using a pre-set first experience library; The reference zero point and the machining control parameters at different wear stages are recorded using a pre-set second experience library to achieve logical isolation of control experience at different wear stages.

4. The adaptive control method for a bit-machining machine tool according to claim 3, characterized in that, Upon detecting a tool change event, the process involves, based on the isolated stored parameters, calling up machining control parameters and a reference zero point that match the state of the changed tool as initial machining parameters, including: After identifying a tool change event, a matching wear stage is retrieved from the first experience base based on the state of the replaced tool. The reference zero point and machining control parameters corresponding to the wear stage are retrieved from the second experience database as initial machining parameters.

5. The adaptive control method for a bit-machining machine tool according to claim 1, characterized in that, The extraction of trend reference components characterizing the evolution trend of tool wear state and transient disturbance components characterizing instantaneous machining deviations based on the process state characteristics includes: The process state characteristics in the continuous processing cycle are smoothed to obtain the trend baseline component; The transient disturbance component is obtained by subtracting the trend reference component from the trend reference component updated in the previous processing cycle from the trend reference component of the current processing cycle.

6. The adaptive control method for a bit-machining machine tool according to claim 1, characterized in that, The step of mapping the current processing state to the corresponding processing branch based on the comparison result, in order to adjust the processing control parameters and reference zero point in the processing cycle, includes: When all dimensions in the trend reference component are less than the stage evolution threshold, and at least one dimension in the transient disturbance component exceeds the tolerance boundary, the transient disturbance component is used as a deviation input, and a preset control algorithm is used to calculate the correction compensation amount for the feed rate or cutting parameters, while keeping the reference zero point unchanged. When at least one dimension of the trend reference component is greater than or equal to the stage evolution threshold, and all dimensions of the transient disturbance component are within the tolerance boundary, the current processing control parameters are maintained unchanged, and the current trend reference component is identified as the new reference zero point. At the same time, the tolerance boundary is adjusted by shifting according to the offset of the reference zero point. When all dimensions in the trend baseline component are less than the stage evolution threshold, and all dimensions in the transient disturbance component are within the tolerance boundary, the current processing control parameters, the reference zero point, and the tolerance boundary are maintained unchanged, and the process state characteristics of the next processing cycle are continuously collected. When at least one dimension of the trend baseline component is greater than or equal to the stage evolution threshold, and at least one dimension of the transient disturbance component exceeds the tolerance boundary, the tool is determined to be severely worn, a tool failure alarm signal is output, the machining cycle is paused, and the machining control parameters and the reference zero point in the machining cycle are synchronously sealed.

7. The adaptive control method for a bit-machining machine tool according to claim 1, characterized in that, The step of mapping the current processing state to the corresponding processing branch based on the comparison result, in order to adjust the processing control parameters and reference zero point in the processing cycle, further includes: When the trend reference component exhibits a continuous drift trend and the transient disturbance component exceeds the tolerance boundary, the current trend reference component is identified as the new reference zero point. The tolerance boundary is adjusted by shifting according to the offset of the reference zero point. Based on the adjusted tolerance boundary, the degree of exceedance of the transient disturbance component is re-evaluated and the processing control parameters are corrected.

8. The adaptive control method for a bit-machining machine tool according to claim 1, characterized in that, The method further includes: When the amplitude of at least one dimension of the transient disturbance component exceeds a preset multiple of the tolerance boundary in a single processing cycle, and falls back to the tolerance boundary in a subsequent preset number of processing cycles, the current processing state is determined to be an occasional transient impact. When the current processing state is determined to be the occasional transient impact, the process state characteristics of the current processing cycle are removed from the calculation of the trend reference component.

9. An adaptive control method for a bit-machining machine tool according to claim 8, characterized in that, The method further includes: The process state characteristics of a preset number of processing cycles are cached; When the amplitude of at least one dimension of the transient disturbance component in the current processing cycle exceeds a preset multiple of the tolerance boundary and does not fall back within a preset time, calculate the mean and variance of the cached process state characteristics. If the values ​​of all dimensions of the variance are lower than the preset steady-state determination threshold, and the value of at least one dimension of the mean is higher than the trend benchmark component, then the current processing state is determined to be a step state caused by tool chipping. When the current machining state is determined to be a step state caused by tool breakage, the current trend reference component is forcibly overwritten using the mean value.

10. An adaptive control system for a bit-making machine tool, for performing the steps of the method according to any one of claims 1 to 9, characterized in that, include: The feature acquisition module is used to acquire process state features in the processing cycle; The feature decomposition module is used to extract, based on the process state features, a trend reference component characterizing the evolution trend of tool wear state, and a transient disturbance component characterizing instantaneous machining deviation. The state comparison module is used to compare the trend baseline component with a preset stage evolution threshold and to compare the transient disturbance component with a preset tolerance boundary. The parameter adjustment module is used to map the current machining state to the corresponding processing branch based on the comparison results, so as to adjust the machining control parameters in the machining cycle and the reference zero point for quantifying the tool wear state. The storage and recall module is used to isolate and store the adjustment records of the machining control parameters and the reference zero point according to the tool wear stage, so that after a tool change event is detected, the machining control parameters and reference zero point that match the state of the replaced tool can be recalled as the initial machining parameters based on the isolated stored parameters.

Images