A decoupled monitoring system for CNC lathe tool degradation and thermal drift and its online tool setting device

By using a decoupled monitoring system for tool degradation and thermal drift on CNC lathes, and by utilizing load measurement and material coefficient calibration of the cutting window group, the problem of inconsistent measurement results caused by tool degradation and thermal drift is solved. This improves tool life management and the accuracy of online tool setting, and ensures the stability of the cutting process.

CN123103447APending Publication Date: 2026-11-17SHENZHEN SHUTE ZHIJIE MASCH CO LTD
View PDF 4 Cites 0 Cited by

Patent Information

Application Number
CN202611634908.7
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-10-21
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

In the continuous machining process of CNC lathes and Swiss-type lathes, inconsistent measurement results caused by tool degradation and thermal drift affect tool life management and the accuracy of online tool setting. Existing technologies are unable to effectively decouple these two effects.

Method used

A decoupled monitoring system for tool degradation and thermal drift of CNC lathes is adopted. By combining the tool setting unit, the acquisition unit and the processing unit, and using the load measurement and material coefficient calibration of the cutting window group, as well as the calibration load model of the shearing term and the cutting edge term, the system can accurately monitor and control the amount of tool degradation.

Benefits of technology

This achieves effective decoupling of tool degradation and thermal drift, improves the accuracy of tool life management and the reliability of online tool setting, reduces the chance of incompatible observation tool setting updates, and ensures the reliability of tool condition history and the stability of the cutting process.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123103447A_ABST
    Figure CN123103447A_ABST
Patent Text Reader

Abstract

This invention relates to CNC monitoring and compensation technology, and discloses a decoupled monitoring system for CNC lathe tool degradation and thermal drift, and its online tool setting device. The system uses a cutting window group, associated with the chip thickness of the cutting edge, to change and then recover according to a program segment. Measurements are performed on the non-cutting reference surface, the cutting edge, and the non-cutting reference surface before and after each window group. After offset restoration, reference drift correction, and differential elongation correction, a corresponding post-group state degradation interval is formed. Within the intersection of the two intervals, candidates whose loads can be explained by the same material coefficient through shear and cutting edge terms are retained, forming a feasible degradation set. The degradation state and reliability are updated when the loads at the beginning and end satisfy the repeatability error bound and the set is not empty; otherwise, it is frozen and retested. Offset correction, spare tool switching, or shutdown retesting are performed in conjunction with the lower bound of the tool life. The device employs an online tool setting probe, temperature and vibration sensors, and signal acquisition components to limit the misattribution of degradation updates to material and thermal state changes.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of CNC machine tool operation monitoring and tool compensation control technology, and in particular to a decoupling monitoring system for CNC lathe tool degradation and thermal drift and its online tool setting device. Background Technology

[0002] During continuous machining on CNC horizontal lathes and Swiss-type lathes, the cutting edge position and cutting load are simultaneously affected by tool degradation, the thermal state of the machine tool and tool, the workpiece material, and the clamping and feed conditions. Tool life management can accumulate cutting time and trigger tool changes, and online tool setting can measure the cutting edge position. However, the accumulated time and actual degradation are not one-to-one, and changes in cutting edge position may also include common thermal displacement and thermal elongation of the tool relative to the tool holder.

[0003] Chinese invention patent application CN113601264A discloses a scheme for determining the flank wear state of a cutting tool based on variable feed rate trial cutting, evaluating wear by changing trial cutting conditions and analyzing feedback signal differences. Chinese invention patent application CN114102260A discloses a scheme for monitoring tool wear under varying operating conditions by combining CNC commands, current, and vibration. The above disclosures demonstrate that the relationship between cutting conditions and load characteristics can be used for tool condition analysis.

[0004] US Patent Publication No. US4120196A discloses a device for detecting tool wear by measuring the relationship between the cutting surface and a non-wearing reference surface; European Patent Application Publication No. EP1327495A1 discloses a scheme for measuring the cutting edge based on a reference position and using it for coordinate correction. In a continuous machining environment on a lathe, when geometric measurement and load monitoring are used together for control, it is also necessary to address the issue of whether the two types of observations correspond to the same cutting edge state.

[0005] Material variations not only affect the load variation with chip thickness but may also influence edge friction and extrusion response; thermal state changes simultaneously affect both tool setting coordinates and actual cutting geometry. If degenerate interpretations are assigned to the geometric and load results separately, without constraints on common states and observation consistency, contradictory observations may be incorporated into the life history, and bias corrections may mask existing physical degradation. Therefore, it is necessary to identify, within the calibrable machining range, whether observations allow for the formation of common degenerate states and to prevent incompatible observations from continuing to drive tool setting updates. Summary of the Invention

[0006] This invention provides a decoupled monitoring system for CNC lathe tool degradation and thermal drift, and its online tool setting device, to solve the problem that when tool degradation, thermal state, and material changes jointly affect the measurement results, incompatible observations are still used for degradation updates and tool setting control.

[0007] To address the aforementioned technical problems, this invention provides a decoupled monitoring system for CNC lathe tool degradation and thermal drift, comprising a tool setting unit, a data acquisition unit, and a processing unit connected to a CNC system. The data acquisition unit associates cutting window groups and window loads of the same cutting edge on the same workpiece according to program segments. The chip thickness of the cutting window group is first changed and then restored. The tool setting unit sequentially measures the non-cutting reference surface on the tool holder supporting the cutting edge, the cutting edge, and the non-cutting reference surface before and after the cutting window group, restoring the measured coordinates to the same offset reference and deducting the interpolation drift and temperature drift of the measured coordinates of the non-cutting reference surface. The differential elongation calibrated by the degree forms a corresponding set of cutting edge degradation intervals relative to the new tool reference. The processing unit selects candidate degradation amounts within the intersection of the cutting edge degradation intervals, and uses the material coefficient to simultaneously participate in the calibration load model of the shear term and the cutting edge term. It retains the candidate degradation amounts that can explain the load of each window within the load error bound with the same material coefficient, forming a feasible degradation set. Only when the load difference between the first and last windows of the cutting window group satisfies the repeatability error bound and the feasible degradation set is not empty, the degradation state and confidence level are updated with the feasible degradation set and the tool setting control signal is output. Otherwise, the degradation state is frozen and a retest is requested.

[0008] Furthermore, the acquisition unit uses program version, program segment number, tool edge identifier, spindle channel, and clamping status to form a stage identifier. According to the stage identifier, cutting time, spindle load, servo current, vibration, temperature, and online measurement deviation are mapped to the normalized cutting process within the program segment. Each window in the cutting window group has the same stage identifier. For Swiss-type lathes, the guide sleeve mode is included in the stage identifier, and the main and auxiliary spindle handover, tool change, and parallel cutting intervals that cannot be separately attributed to tool edge loads are excluded.

[0009] Furthermore, the non-cutting reference surface and the cutting edge are located on the same locked tool holder. The tool setting unit records the coordinates, time, and effective offset value of each measurement group under the same measurement direction and trigger speed. The measurement coordinates are restored using the effective offset value. The position of the non-cutting reference surface at the measurement time of the cutting edge is obtained by interpolation according to the time ratio between the measurement time of the cutting edge and the measurement time of the two non-cutting reference surfaces. After the displacement change of the cutting edge relative to the position of the non-cutting reference surface is corrected by the differential elongation, the cutting edge position degradation amount relative to the new tool reference is determined along the cutting edge retraction direction. The probe repeatability error, interpolation error, offset restoration error, differential elongation calibration error, new tool reference error, and intra-group retraction change error are propagated into the cutting edge position degradation interval.

[0010] Furthermore, the differential elongation is the thermal elongation of the cutting edge relative to the non-cutting reference surface. It is calibrated by the displacement change of the cutting edge relative to the non-cutting reference surface, the tool holder temperature, and the near-cutting temperature of the tool during heating and cooling processes when no cutting occurs. The heating and cooling processes are respectively used to establish calibration relationships. The processing unit selects the calibration relationship according to the direction of temperature change. If the temperature exceeds the calibration range, the direction of temperature change cannot be determined, or the difference between the coordinates of two measurements of the non-cutting reference surface exceeds the allowable drift, the processing unit refuses to allow the cutting edge degradation interval to participate in the update and requests a retest.

[0011] Further, the cutting window group includes a starting window, at least one intermediate window, and a recovery window arranged in sequence. The starting window, the intermediate window, and the recovery window are located in adjacent equal depth of cut regions of the same workpiece, and have the same cutting speed, cutting width, and cooling state. The starting window and the recovery window use the same undeformed chip thickness, while the intermediate window uses a different undeformed chip thickness than the starting window. The processing unit deducts the no-load load at the same rotation speed, converts the spindle load into a tangential force per unit cutting width to constitute the window load, and uses servo current and vibration to mark windows that pass steady-state detection as valid windows. When the starting window or the recovery window is not marked as a valid window, no intermediate window is marked as a valid window, the chip thickness difference is lower than the identifiable threshold, or the window load difference between the starting window and the recovery window exceeds the repeatability error limit, the feasible degradation set is set to empty.

[0012] Further, the calibration load model takes the candidate degradation amount, the material coefficient, and the undeformed chip thickness as inputs. The shear term is the product of the material coefficient and the undeformed chip thickness, and the cutting edge term is the calibration response that varies with both the candidate degradation amount and the material coefficient. The shear term and the cutting edge term are added together to obtain the predicted window load. For each candidate degradation amount, the processing unit calculates, within the material coefficient calibration range, the set of material coefficients that makes the difference between the predicted window load and each window load fall within the load error bound formed by the measurement error, model calibration error, and intra-group shrinkage variation propagation error. Candidate degradation amounts whose intersection of each set of material coefficients is not empty are included in the feasible degradation set, and the correlation between each candidate degradation amount in the feasible degradation set and the intersection of the material coefficients is saved. When all candidate degradation amounts do not satisfy the load error bound, the feasible degradation set is set to empty.

[0013] Further, the processing unit determines the reliability based on the span of the feasible degradation set, the proportion of the effective window, and the measurement aging of the cutting edge degradation interval; when the feasible degradation set is not empty, the feasible degradation set and the cumulative cutting time are stored in the degradation state, and a degradation rate interval is formed by utilizing the change of candidate degradation amount with cumulative cutting time in multiple effective updates, and the lower bound of the remaining life is determined based on the upper bound of the feasible degradation set, the upper bound of the degradation rate interval, and the preset tool change degradation amount threshold; when the feasible degradation set is empty, the previously effective degradation state is maintained; when the feasible degradation set is empty, the upper bound of the degradation rate interval cannot be determined, or the machining conditions exceed the applicable range of the degradation rate estimation, the reliability is reduced and a life unavailable state is output.

[0014] Further, the tool setting control signal includes an offset correction command, a spare tool switching command, or a stop and retest command. The stop and retest command takes precedence over the spare tool switching command, and the spare tool switching command takes precedence over the offset correction command. When the reliability meets the preset conditions and the upper bound of the feasible degradation set is less than the preset tool change degradation threshold, the processing unit generates the limited offset correction command based on the thermal drift measured by the tool setting and the online measurement deviation. When the lower bound of the remaining lifespan is less than the expected cutting time of the next machining task or the upper bound of the feasible degradation set reaches the preset tool change degradation threshold, the spare tool switching command pointing to the spare tool that has passed the tool setting verification is generated. When the reliability does not meet the preset conditions, the lifespan becomes unavailable, the online measurement deviation still exceeds the limit after offset correction, or the spare tool verification fails, the stop and retest command is generated. The processing unit stores the offset values ​​before and after offset correction and the effective time in association with the measurement coordinates. Offset correction does not reset the degradation state. After the tool is replaced, a new tool reference and degradation state are established.

[0015] This invention also provides an online tool setting device, including an online tool setting probe, a tool holder equipped with a cutting edge and a non-cutting reference surface, a temperature sensor, a signal acquisition component, and a processor connected to a CNC system; the signal acquisition component associates the cutting window group and window load of the cutting edge on the same workpiece according to a program segment, and the chip thickness of the cutting window group is changed and then restored; the processor controls the online tool setting probe to sequentially measure the non-cutting reference surface, the cutting edge, and the non-cutting reference surface before and after the cutting window group, restore the measurement coordinates to the same offset reference, and subtract the non-cutting reference surface. The interpolation drift of the measured coordinates and the differential elongation calibrated according to the data of the temperature sensor are used to form the cutting edge degradation interval of the corresponding group state relative to the new tool reference. The processor selects candidate degradation amounts within the intersection of the cutting edge degradation intervals, and uses the material coefficient to simultaneously participate in the calibration load model of the shear term and the cutting edge term. It retains the candidate degradation amounts that can explain the load of each window within the load error boundary with the same material coefficient, forming a feasible degradation set. When the load difference between the first and last windows of the cutting window group satisfies the repeatability error boundary and the feasible degradation set is not empty, the tool setting correction permission is output according to the feasible degradation set; otherwise, a retest is requested.

[0016] Furthermore, it also includes a vibration sensor fixed to the tool holder, and the temperature sensors are respectively arranged on the tool holder and the near-cutting edge of the tool. The signal acquisition component identifies the unsteady window in the cutting window group according to the signal of the vibration sensor. When an unsteady window exists, the processor sets the feasible degenerate set to empty. The processor only allows group measurement when the spindle stops, the tool holder is locked, and the online tool setting probe is in position signal is valid. If the effective bias value changes or the measurement sequence is incomplete during the group measurement, the group measurement is discarded, and the machining feed prohibition state is lifted after the confirmation signal of the online tool setting probe exiting the measurement position is valid.

[0017] This invention limits candidate degradation amounts by defining the geometric allowable range before and after the window group, and requires material coefficients to jointly explain the load of each window in the shear and cutting edge terms, thus constraining geometric degradation and material response under the same candidate state. Load verification after restoring to the same chip thickness provides a check for short-term state consistency. Reference surface interpolation and differential elongation correction handle common displacement and relative thermal elongation of the tool, respectively, while offset restoration maintains comparability of physical states before and after tool compensation changes. Observations that cannot meet common constraints are frozen and updated, and compensation, switching, and retesting are constrained by confidence and conditional life results, reducing the chance of incompatible observations accumulating into the tool state history. Attached Figure Description

[0018] Figure 1 This is a schematic diagram showing the composition and control connection of a decoupling monitoring system for tool degradation and thermal drift on a CNC lathe.

[0019] Figure 2 This is a schematic diagram showing the arrangement of the probe, tool holder, non-cutting reference surface, and sensors in an online tool setting device.

[0020] Figure 3 This is a schematic diagram illustrating the timing and constraint relationships between the preceding and following group tool setting measurements, the cutting window group, and the feasible degenerate set. Detailed Implementation

[0021] This invention relates to tool condition monitoring and tool setting control for CNC horizontal lathes and Swiss-type lathes. Taking continuous longitudinal external turning of metal bars as an example, the tool is mounted on a tool holder, which is locked to a turret or tool chuck mechanism. The CNC system controls the spindle and feed axis. The main implementation operates within the calibration range corresponding to the same material grade, heat treatment state, and insert geometry; tool edge deterioration is characterized by tool edge retraction in a selected measurement direction. This retraction amount and the width of the flank wear band represent different geometric objects. Load changes caused by edge dulling, built-up edge, and chipping are handled by the mismatch conditions described later.

[0022] like Figure 1 As shown, the CNC lathe tool degradation and thermal drift decoupling monitoring system of the present invention includes a tool setting unit, a data acquisition unit, and a processing unit connected to the CNC system. The tool setting unit is connected to the probe trigger interface of the CNC system, the data acquisition unit reads the machining stage and sensor signals, and the processing unit receives the tool setting measurement results and window load. The processing unit uses the commonly allowed cutting edge state and material response to determine whether the current observation can enter the degradation history, and then transmits the state results to the tool compensation and tool change interface of the CNC system.

[0023] like Figure 2 As shown, the online tool setting device includes an online tool setting probe, a tool holder with a cutting edge and a non-cutting reference surface, a temperature sensor, a signal acquisition component, and a processor connected to the CNC system. The online tool setting probe is mounted on a measuring arm with extension and retraction confirmation switches; the tool is fixed to the tool holder with clamping screws and a positioning surface; the non-cutting reference surface is located on a part of the tool holder that does not contact the workpiece; the cutting edge belongs to the clamped tool. The signal acquisition component transmits trigger coordinates and sampling records to the processor, which then issues measurement movement, feed restriction, tool setting correction permission, and retest requests via the CNC interface.

[0024] When the online tool setting device is connected to the system, the tool setting unit consists of an online tool setting probe, a measuring arm control circuit, and a geometric calculation program; the acquisition unit consists of a signal acquisition component and its stage-related program; and the processing unit consists of a joint identification program and a control program in the processor. The geometric calculation program and the joint identification program can run on the same processor, exchanging data through records with tool edge markings and measurement group numbers; the division of program execution does not change the measurement, acquisition, and processing relationships undertaken by each unit.

[0025] The non-cutting reference surface is a planar measurement area adapted to the probe's approach direction. Its normal is consistent with the selected measurement direction of the cutting edge and is located where the probe can avoid contact with the workpiece and fixture. During assembly, the positional relationship between the non-cutting reference surface and the tool mounting and positioning surface is fixed, and the same tool holder is used so that both share the overall displacement of the tool holder. The cutting edge still has a differential elongation relative to the non-cutting reference surface caused by the tool extension length and temperature gradient, which is calibrated separately. Before each measurement, chips and adhering coolant in the measurement area are removed, and the tool holder locking feedback is checked; the geometric datum is re-established after re-clamping the tool or adjusting the non-cutting reference surface.

[0026] The cutting window group is a set of cutting windows associated with the same cutting edge according to the cutting process. The feasible degradation set consists of candidate degradation values ​​that simultaneously satisfy geometric measurement and load constraints of each window. The online tool setting device also includes a vibration sensor fixed to the tool holder. The temperature sensors are respectively arranged on the tool holder and the near-edge portion of the tool. The signal acquisition component identifies the non-steady-state windows in the cutting window group based on the signals from the vibration sensors. When a non-steady-state window exists, the processor clears the feasible degradation set. The vibration sensor is installed on the outer surface of the tool holder near the tool clamping point, and the signal line is connected to the signal acquisition component via a shielded cable. The temperature measurement point of the tool holder is located near the reference surface, and the temperature measurement point of the near-edge portion of the tool is located on the tool surface that does not participate in cutting and does not interfere with the proximity of the probe. The near-edge temperature measurement can use an embedded thermocouple or a fixedly installed non-contact temperature measurement element. When using a non-contact method, the errors caused by emissivity, field obstruction, and coolant coverage are included in the temperature calibration package. The device adopts a whole-group steady-state rejection method. If any predetermined window becomes unstable, the whole group of data is reacquired.

[0027] The acquisition unit uses program version, program segment number, tool edge identifier, spindle channel, and clamping status to form stage identifiers. Cutting time, spindle load, servo current, vibration, temperature, and online measurement deviation are mapped to the normalized cutting process within the program segment according to these stage identifiers. Each window within the cutting window group has the same stage identifier. The program version is taken from the CNC program verification identifier, the program segment number is taken from the currently executing machining segment, and the tool edge identifier includes the tool position number and the lifecycle number of the current tool edge at that tool position. The spindle channel distinguishes the spindle drive source, and the clamping status distinguishes the clamping position and clamping method. Cutting time only accumulates the time when the tool edge actually participates in cutting and can be attributed to that tool edge; idle travel, tool change, and probe measurement time are not considered cutting exposure.

[0028] For Swiss-type lathes, the guide sleeve pattern is included in the stage identifier, and parallel cutting intervals during spindle and counterspindle transitions, tool changes, and when the load cannot be separately assigned to the cutting edge are excluded. For Swiss-type lathes where both the spindle and counterspindle can drive the workpiece, a cutting window group is only established when it is confirmed that the workpiece is driven by a designated spindle, the guide sleeve pattern remains unchanged, and the spindle load can be assigned to the current cutting edge. The servo synchronization current during spindle and counterspindle transitions cannot be used as a single-edge cutting load; after exclusion, its time stamp is retained to avoid piecing together the data on both sides of the transition segment into a continuous valid window.

[0029] The acquisition unit uses the same hardware timing reference or a clock calibrated for deviation to record each signal. The probe trigger edge, spindle load, and servo current retain their original timestamps. Vibration is first anti-aliasing filtered at the acquisition end, and then root mean square and peak characteristics are formed according to the same window. Temperature is maintained or interpolated according to the actual sampling time, while retaining the time from the nearest original sample. Window statistics are only allowed when the sampling period and synchronization error are sufficiently small relative to the shortest steady-state window; when there is packet loss, saturation, non-finite sampling values, time reversal, or synchronization error exceeding the calibrated upper limit, the corresponding window is marked as invalid.

[0030] To ensure that signals with different sampling rates correspond to the actual tool positions within the same program segment, the acquisition unit uses the arc length of the tool position along the effective cutting path of that segment to determine the normalized cutting progress:

[0031]

[0032] Among them, the normalized cutting process It is a dimensionless quantity; the current arc length Effective cutting start arc length and the arc length of the endpoint All positions are obtained through CNC interpolation mapping, with units in millimeters, and the endpoint arc length is greater than the starting arc length. Each signal obtains a normalized cutting progress based on the tool position corresponding to the timestamp. Processes outside the range of 0 to 1, or those segments lacking effective cutting length, are not included in the alignment. This mapping maintains the positional identity of different feed windows, forming window groups before comparing window loads.

[0033] The acquisition unit associates cutting window groups and window loads on the same workpiece and the same cutting edge according to program segments. The chip thickness of the cutting window group is changed and then restored. The cutting window group includes a starting window, at least one intermediate window, and a recovery window arranged in sequence. The starting window, the intermediate window, and the recovery window are located in adjacent equal depth of cut regions of the same workpiece, with the same cutting speed, cutting width, and cooling state. The starting window and the recovery window use the same undeformed chip thickness, while the intermediate window uses a different undeformed chip thickness than the starting window. In the same relatively long straight-line turning program segment, the processing unit sequentially applies the initial feed, the changed feed, and the restored initial feed through the feed rate interface allowed by the CNC system. The machining segment number remains unchanged, and the transition part of the rate change and servo response is not included in the window; each window starts after the actual feed, speed, and depth of cut have stabilized. Adjacent areas have machining allowances that can be used for diagnosis, and the diagnostic feed and duration are both within the tool's allowable cutting range.

[0034] For longitudinal turning where the main cutting edge is approximately straight and the influence of the tool tip radius can be accounted for, the feed per revolution and radial depth of cut are converted to the undeformed chip thickness and cutting width used in the load model:

[0035]

[0036] Among them, the thickness of the undeformed chip and cutting width The unit is millimeters; feed per revolution The unit is millimeters per revolution, radial depth of cut. The unit is millimeters, principal deflection angle Determined by tool mounting geometry, sine calculation The angle input is in radians, with a sine value greater than 0. The calculated thickness is used to distinguish between the starting, intermediate, and recovery windows, and the width is used to normalize the load. When the tool tip radius is dominant, the depth of cut is close to zero, or the cutting geometry cannot be determined, the effective thickness and width calibrated by geometry are used; if there is no corresponding calibration package, this identification is not performed.

[0037] The processor allows group measurements only when the spindle is stopped, the tool holder is locked, and the online tool setting probe is in position signal is valid. If the effective offset value changes or the measurement sequence is incomplete during group measurements, the group measurement is discarded, and the machining feed prohibition state is released after the online tool setting probe exits the measurement position confirmation signal is valid. Before entering group measurements, the CNC system first ends cutting and moves the tool to a position where it will not collide with the workpiece, then stops the spindle, keeps the tool holder locked, and extends the measuring arm. The processor latches three allow signals with the same measurement group number; if the signal is withdrawn or times out, the machining feed prohibition remains and the results of this group are cancelled. After the measurement is completed, the probe contact surface is withdrawn first, and then the measuring arm is retracted. Exit confirmation only releases the feed prohibition caused by the probe position; other shutdown conditions are still handled according to their own reset conditions.

[0038] The non-cutting reference surface and the cutting edge are located on the same locked tool holder. The tool setting unit records the coordinates, time, and effective offset value of each measurement under the same measurement direction and trigger speed, and uses the effective offset value to reconstruct the measurement coordinates. The measurement direction is taken as the physical direction from the non-cutting reference surface to the new cutting edge; the cutting edge retracts in the opposite direction as positive degradation. The same direction and trigger speed are used to reduce changes in probe pre-stroke, contact stiffness, and backlash. Each trigger saves the coordinate type, length unit, and effective tool compensation version to avoid subtracting coordinates with different representations.

[0039] To eliminate the impact of CNC tool compensation rewriting on geometric history, the interface readings for each trigger are first converted to physical measurement coordinates under the same offset reference:

[0040]

[0041] Among them, unified measurement coordinates Interface trigger coordinates The final length unit is millimeters; offset and coordinate transformation records. Includes the tool offset, coordinate system, and diameter or radius representation for the current operation; coordinate restoration function. Calibrate according to the coordinate definition of the controller used. If the interface is defined as the reading equal to the physical coordinate minus the offset, then the sum of the two is used for restoration; if the original machine tool coordinates without tool offset have already been read, then the tool offset is not superimposed again. When the radial coordinates of the lathe are displayed using the diameter, first convert them to the actual displacement in the radial direction, and then unify the approximate direction sign.

[0042] The tool setting unit sequentially measures the non-cutting reference surface on the tool holder carrying the cutting edge, the cutting edge, and the non-cutting reference surface before and after the cutting window group. It restores the measurement coordinates to the same offset reference, subtracts the interpolation drift of the measurement coordinates of the non-cutting reference surface and the differential elongation due to temperature calibration, and forms the corresponding edge position degradation range relative to the new tool reference. The two non-cutting reference surface measurements, clamped while measuring the cutting edge, are used to estimate the common reference position corresponding to the moment the cutting edge is triggered, rather than directly using the previous reference surface position for the subsequent cutting edge position. The new tool reference is acquired when the cutting edge is not participating in cutting, the tool holder is clamped, the measurement direction is fixed, and the temperature is at the reference state. It saves the reference surface position, the distance of the cutting edge relative to the reference surface, and the error boundary.

[0043] For a complete measurement sequence of reference surface, cutting edge, and reference surface, linear interpolation is performed according to the trigger time to obtain the position of the non-cutting reference surface at the time of cutting edge measurement:

[0044]

[0045] Among them, the interpolation ratio Dimensionless; First non-cutting reference surface trigger time Blade trigger moment and the triggering time of the second non-cutting reference surface Both use the same clock, in seconds, and strictly increase in that order; the reference plane coordinates after the two reconstructions. , and the position of the interpolation reference surface The unit is millimeters. This group is discarded if the time span is zero, the blade trigger is not between two reference surface triggers, or either trigger is missing. The interpolation result is used to subtract common displacements during the measurement period; the interpolation residual is separately included in the geometric error.

[0046] The differential elongation is the thermal elongation of the cutting edge relative to the non-cutting reference surface. It is calibrated by measuring the displacement change of the cutting edge relative to the non-cutting reference surface, the tool holder temperature, and the near-edge temperature of the tool during heating and cooling processes without cutting. Calibration relationships are established for heating and cooling respectively. During calibration, a controlled thermal process without cutting is used. A temperature combination covering the usage range is formed using a temperature-controlled heat source and machine tool idle operation. Reference surface, cutting edge, and reference surface measurements are repeatedly performed on the same new tool. The applicable range of tool holder temperature, near-edge temperature, temperature change rate, and waiting time is saved for each heating or cooling branch. The change in relative distance between the new tool and the cutting edge is used as the differential elongation calibration value. Two branches are fitted separately, and the residual envelope is determined using the temperature path not involved in the fitting. Measurements after cutting must enter the thermal process covered by the envelope; internal temperature gradients not reflected by the measurement points are limited by the residual and waiting conditions.

[0047] Under the condition that the temperature range is small and the calibration residuals meet the allocated error budget, the following differential elongation relationship is adopted for each thermal process branch:

[0048]

[0049] Among them, differential elongation The unit is millimeters; tool holder temperature Temperature near the cutting edge of the tool The reference temperature was obtained from the corresponding measuring point. Corresponding to the new blade reference, the temperature unit is degrees Celsius; branch intercept The unit is millimeters, and the temperature coefficient of the tool holder is... Temperature coefficient near the cutting edge The unit is millimeters per degree Celsius. Each coefficient is obtained from the heating or cooling records and is read along with the selected branch. If the linear residual is too large, a piecewise bounded interpolation table with the same temperature input is used, still outputting the differential elongation and corresponding error bound, and prohibiting extension outside the calibration range.

[0050] The processing unit selects the calibration relationship based on the direction of temperature change. If the temperature exceeds the calibration range, the direction of temperature change cannot be determined, or the difference between the coordinates of two non-cutting reference surface measurements exceeds the allowable drift, the unit rejects the cutting edge degradation interval from participating in the update and requests a retest. The direction of temperature change is determined through continuous raw temperature samples and dead zones. When in a stable dead zone, a branch is maintained only if there is a traceable branch from the previous thermal process and the stabilization time meets the calibration conditions for that branch. If there is no branch record at startup, the trend of change at two measurement points does not match any calibration path, or the measurement waiting conditions are not met, the unit treats the situation as if the direction of temperature change cannot be determined. The allowable drift is determined by the linear interpolation residual requirement within the measurement span, and the measurement span is also limited. Even a small difference between the coordinates at both ends cannot eliminate the residual limitation on nonlinear drift.

[0051] By interpolating the time ratio between the measurement time of the cutting edge and the measurement time of the non-cutting reference surface between two measurement times, the position of the non-cutting reference surface at the measurement time of the cutting edge is obtained. After the displacement change of the cutting edge relative to the position of the non-cutting reference surface is corrected by the differential elongation, the amount of cutting edge degradation relative to the new tool reference is determined along the cutting edge retraction direction. The probe repeatability error, interpolation error, offset restoration error, differential elongation calibration error, new tool reference error, and intra-group retraction change error are propagated into the cutting edge degradation interval. The measurement coordinates of the cutting edge and the position of the interpolated reference surface are first subtracted to form a relative distance, and then the reversible differential elongation is subtracted from the relative distance. This sequence processes the overall movement of the tool holder and the thermal elongation of the tool relative to the tool holder separately, so that the tool compensation history only participates in coordinate restoration and is not used as the basis for resetting the degradation state to zero.

[0052] To express the reduction in relative distance as a positive edge degeneration, the following calculations are performed on the tool edge coordinates that have been standardized in the same direction:

[0053]

[0054] Wherein, the distance between the current cutting edge and the non-cutting reference surface. Unified blade measurement coordinates Position of the interpolation reference surface at the blade tip moment Reference distance for new knife Differential elongation Degeneration of the cutting edge All measurements are in millimeters. The reference distance for the new blade is obtained from the initialization of the blade's lifespan; the differential elongation is calculated based on the current thermal process branch. Positive degradation indicates that the relative distance between the corrected blade edges has shortened. Negative results, if within the measurement error range, retain their bounded uncertainty and intersect with the non-negative physical degradation domain; if the entire interval is negative, the measurement is repeated according to the baseline, built-up edge, or model mismatch.

[0055] The blade position degradation range is formed using bounded error propagation. Probe repeatability is determined by the discrete range of repeated triggering at the same direction and speed. The reference surface triggering error is propagated according to interpolation weights. The interpolation error is determined by the curvature boundary of the reference drift and the longest measurement span during calibration. The offset restoration error includes coordinate quantization and uncertainty at the effective time. The thermal calibration error is taken as the residual envelope of the corresponding temperature branch. The new blade reference error is retained independently. When there is no independent random error basis, the above absolute error boundaries are added together, without using the sum of squares to reduce the error. The new blade reference and offset errors shared by the two measurements maintain the same source in the joint processing, and the two observations are jointly propagated according to the shared error source.

[0056] The duration of a window group is constrained by a short-term degradation change budget. The maximum intra-group regression change is determined by the upper bound of the degradation rate covering the diagnostic conditions in calibration and the longest group duration. This change is included in the error budget of the pre-group geometric measurement corresponding to the post-group state, and the difference between the cutting period and the post-group state is propagated to the load model error. If a tighter effective rate upper bound is available in production, this budget can be shortened; if there is no coverage upper bound, controlled diagnostic calibration is performed first to obtain the intra-group regression change budget covering the diagnostic conditions before forming the usable interval. The intersection of the pre-group and post-group edge position degradation intervals is constrained to the same post-group degradation candidate. If the two intervals do not intersect, the group is rejected. No additional translation beyond the above time budget is performed before the intervals intersect.

[0057] The allowable range of geometric measurements before and after the window group, together with the non-negative degenerate calibration domain, are used to limit the candidate search range:

[0058]

[0059] Among them, candidate geometric range Degenerative range of the front edge position Degenerative range of the combined rear edge position All are collections of millimeter-scale measurements; upper limit of degradation calibration. The coverage area is determined by the blade used and the calibration load model, in millimeters. The intersection operation retains quantities that simultaneously belong to each range, with 0 indicating the starting point of the new blade geometry degradation. The allowable ranges for preceding and following measurements include their respective measurement errors and corresponding time budgets; when the candidate geometry range is empty, subsequent load calculations do not generate usable degradation updates.

[0060] The following uses a set of input settings to illustrate the geometric calculations. The reference distance for the new blade is set to 5.000 mm. The three trigger times are 0 seconds, 1 second, and 2 seconds respectively. The unified coordinates of the two reference surfaces are 10.000 mm and 10.004 mm, the unified coordinate of the blade edge is 14.990 mm, and the differential elongation is 0.008 mm. The interpolated reference surface position is 10.002 mm, the relative distance between the blade edges is 4.988 mm, the distance after differential elongation correction is 4.980 mm, and the blade position degradation is 0.020 mm. If the intra-group shrinkage variation budget for the corresponding post-group state is taken into account, the total error limit before the group is 0.002 mm, then the blade position degradation range before the group is 0.018 to 0.022 mm; when the measured value after the group is 0.021 mm and the total error limit is 0.002 mm, the intersection of the two ranges is 0.019 to 0.022 mm. These values ​​are used to explain the calculation relationships, and the error limits of specific machine tools are determined by the aforementioned calibration process.

[0061] The processing unit deducts the no-load load at the same speed, converts the spindle load into a tangential force per unit cutting width to form the window load, and marks the windows that pass steady-state detection as valid windows using servo current and vibration. When the spindle load is a percentage or current reading, it is first converted into spindle torque according to the motor and transmission calibration relationship of the drive channel; the percentage value cannot be directly used as the tangential force. The no-load record covers the same speed, rotation direction, drive mode, and cooling auxiliary load; the inertial torque during spindle acceleration and deceleration is not included in the steady-state cutting load.

[0062] When the spindle input is already calibrated as torque, the net torque is converted into a window load using the workpiece cutting radius and cutting width:

[0063]

[0064] Among them, window load For the first Tangential force per unit cutting width within an effective window, in Newtons per millimeter; average spindle torque within the window. and no-load torque at the same speed The unit is Newton-meter; cutting radius Meters, cutting width Use millimeters, subscript This indicates the sequence number of the valid windows within the window group. The radius is determined by the machining geometry and available online dimensions, while the width is obtained from the actual depth of cut and tool angle. A window is rejected if the radius or width is not positive, the torque calibration is out of bounds, or the geometric deviation exceeds the error budget.

[0065] Steady-state detection simultaneously checks the variation range of the actual CNC speed and feed, the mean drift and peak value of the servo current, and the root mean square and impact peak value of the tool holder vibration. Each range is formed by forming an envelope from stable cutting records under the same cutting tool and working conditions, and is limited by sampling noise and allowable working condition variations; if the vibration exceeds the envelope, the servo current changes abruptly, the sensor saturates, or the workpiece is eccentric, making it impossible to determine the actual cutting width, the window is not used. The acquisition unit records the predetermined number of windows and the number of valid windows for reliability assessment. The main implementation uses the condition that all three predetermined windows pass; when expanded to multiple intermediate windows, the identity and time sequence of the starting and recovery windows are still retained.

[0066] The calibration load model characterizes effective shear resistance using a material coefficient and includes a shear term and a cutting edge term. The calibration load model takes the candidate degradation amount, the material coefficient, and the undeformed chip thickness as inputs. The shear term is the product of the material coefficient and the undeformed chip thickness, and the cutting edge term is the calibration response that varies with both the candidate degradation amount and the material coefficient. The shear term and the cutting edge term are added to obtain the prediction window load. The material coefficient characterizes the effective shear resistance within the current material and process calibration package, and the cutting edge term includes the new cutting edge response and the additional response that varies with degradation. Material variations also change friction, compression, and plowing at the cutting edge; therefore, the material coefficient is used in both terms, and not all intercept changes are directly interpreted as cutting edge degradation.

[0067] When examining candidate degradation values ​​one by one within geometrically permissible limits, the predicted window load is formed based on the chip thickness of the current window:

[0068]

[0069] Among them, the predicted window load and calibration edge response The unit is Newtons per millimeter; material coefficient The unit is Newtons per square millimeter, and the thickness of the undeformed chip at the window is... and candidate degradation amount The unit is millimeters, and the subscript is... Indicates the valid window number. Calibrates the cutting edge response function. The candidate degradation amount and material coefficient are used as inputs, derived from calibrations under the same material family, blade geometry, speed, depth of cut, and cooling conditions. Each window uses the same candidate degradation amount and the material coefficient to be tested, while the thickness is taken from the actual value of each window.

[0070] When establishing a load calibration package, select material states covering the expected resistance range within the selected material family. Obtain new cutting tools and multiple measurable retraction states for the same type of insert. The retraction states are determined using the aforementioned geometric measurements and corrected according to the same thermal constraints. For each retraction state, use different feed rates to form steady-state windows of different thicknesses. First, estimate the material coefficients by the load variation with thickness, and then determine the cutting edge response by the load after deducting the shear term. The calibration record simultaneously saves the insert geometry, principal cutting edge angle, material grade and heat treatment state, speed, depth of cut, cooling method, and temperature conditions. When the calculated material coefficients under different retraction states are significantly inconsistent, check the uniformity of the samples and the changes in working conditions, and retain the retraction labels determined by geometric measurements. Bilinear calibration should cover at least two distinguishable material resistance states and new cutting tools and non-zero retraction states. Use the material coefficient, the product of the material coefficient and the retraction amount, and the retraction amount as three regression inputs to check if the calibration matrix columns are full rank. If full rank is achieved but sensitive to measurement disturbances, expand the calibrable input differences or narrow the application range, and do not publish coefficient combinations that cannot be stably identified.

[0071] When the cutting edge response within the same calibration package can be covered by the bilinear relationship between the material and degradation, the following directly solvable form can be used:

[0072]

[0073] Among them, the calibrated cutting edge response The output is in Newtons per millimeter, and the candidate degradation amount is... Material coefficient in millimeters Measured per ox per square millimeter; equivalent blade length of a new knife Material conditional degradation coefficient (in millimeters) Dimensionless, with additional degradation response coefficient The coefficients are expressed in Newtons per square millimeter. The three coefficients are jointly obtained from calibration records under different material conditions and different degradation conditions, and are not arbitrarily adjusted with the current identification window. Within the selected range, the load should maintain a recognizable response to degradation, and the denominator of the material coefficients calculated later should be positive; otherwise, the calibration range should be split or this form should be rejected.

[0074] The material state not involved in coefficient calculation is combined with the degradation state to test the predicted residuals. The residual envelope plus the instrument uncertainty forms the model calibration error. The test combination must cover the upper and lower limits of the material coefficients, the upper and lower limits of degradation, and the diagnostic thickness range. If the friction or built-up edge state corresponding to the same material coefficient causes the cutting edge response to be divided into different relationships that cannot be covered by the allowable error, then calibration packages are set up according to material and cooling conditions. Alternatively, a two-dimensional piecewise interpolation table of candidate degradation amounts and material coefficients can be saved, retaining the upper and lower limits of the response and interpolation error for each element; its application is limited to the element boundaries supported by calibration.

[0075] Reference Figure 3 As shown, after the pre-group geometric measurement, thickness change and recovery cutting, and post-group geometric measurement are completed sequentially, the processing unit selects candidate degradation quantities within the intersection of the cutting edge degradation intervals. Using a calibration load model where material coefficients simultaneously participate in both the shear and cutting edge terms, candidate degradation quantities that can explain the load of each window within the load error bound with the same material coefficient are retained, forming a feasible degradation set. Geometric intersection excludes degradation quantities incompatible with direct tool-setting measurements, while the same material coefficient constraint across windows excludes degradation quantities that require separate material coefficients for each window to explain. Both types of observations act on the same candidate, and the updated result is formed by the candidate set that simultaneously satisfies both types of constraints; the average of the estimates from two independent points is not used to replace this set.

[0076] For each candidate degradation, the processing unit calculates a set of material coefficients within the material coefficient calibration range that ensures the difference between the predicted window load and the loads of each window falls within the load error bounds formed by measurement error, model calibration error, and intra-group shrinkage propagation error. Candidate degradations whose intersection with each set of material coefficients is not empty are included in the feasible degradation set, and the correlation between each candidate degradation and the intersection of material coefficients in the feasible degradation set is preserved. The load error bounds correspond to each window and, in addition to torque measurement and no-load subtraction errors, include actual thickness, width, and cutting radius errors, intra-group shrinkage propagation terms, and model calibration residuals. When related sources exist, common error variables are maintained, or conservative absolute bound propagation is used. When thermal drift causes changes in depth of cut, the cutting geometry and its errors are corrected first; if the changes exceed the allowable range, the changes are rejected to prevent the material coefficients from absorbing all depth of cut changes.

[0077] To test whether a given geometric candidate can be explained by the same material response, allowable sets of material coefficients are established for each effective window:

[0078]

[0079] Among them, the set of material coefficients Includes making the first Material coefficients whose residuals within a given window meet the limits; material calibration range. The set of material coefficients covered by the selected calibration package, material coefficients Window load is measured in Newtons per square millimeter. , Edge response function Output and load error limits All measurements are in Newtons per millimeter (N / mm), window thickness. and candidate degradation amount Measured in millimeters. The absolute value represents the difference between the measured load and the predicted load. The colon in the set indicates that the residual condition following it must be met simultaneously. The resulting windows allow the sets to proceed to the next step of common intersection testing.

[0080] Intersect the sets of material coefficients for each window corresponding to the same candidate, and then project the candidates that still have common material coefficients into a feasible degenerate set:

[0081]

[0082] Among them, the set of common material coefficients The set of material coefficients for each effective window The intersection, intersection index Traverse the valid windows of this group; candidate degradation quantity Taken from geometric candidate range Units are millimeters; feasible degenerate set Store candidates that exist in a non-empty set of common material coefficients. An empty set symbol indicates that no quantity satisfies the conditions. The set of common material coefficients is expressed in Newtons per square millimeter, and the feasible degenerate set is expressed in millimeters. Store the correspondence between candidates and the set of common material coefficients so that during review, it is possible to trace how each candidate interprets the entire window.

[0083] For the bilinear cutting edge model described above, within the calibration range where the denominator is positive, the allowable set of material coefficients for each window can be directly represented by the upper and lower bounds:

[0084]

[0085] Among them, the set of material coefficients and material specification range Measured window load in units of Newtons per square millimeter; and load error boundary Candidate degradation per millimeter (in cows); Window thickness and the equivalent blade length of the new knife In millimeters; degradation coefficient Dimensionless, with additional response coefficient In terms of cattle per square millimeter, the subscript is... Indicates the valid window. Square brackets indicate a closed interval including the endpoints. The two fractions are derived from solving the material coefficients using the residual inequality. This formula is not applicable if the denominator is not positive; a calibrated piecewise solution or the output model will be unavailable.

[0086] The processing unit combines the conditions that the lower bound of the material coefficients of each window is not greater than the upper bound of other windows and intersects with the material calibration range. For the bilinear form, all denominators are positive, and the identical quadratic terms cancel each other out after cross-multiplication, resulting in a linear inequality about the candidate degradation quantity. Combining these inequalities with the geometric candidate range yields a continuous feasible range, avoiding the omission of real-valued candidates by only scanning a few grid points. The piecewise model is solved for each calibration unit, and the union of each effective branch is saved; discontinuous parts between branches are not considered feasible quantities. Numerical representation errors are covered by outward rounding of endpoints, and the existence of candidates is still determined by the bounded solution results; only candidate degradation quantities with common material coefficients proven by bounded solution within a piecewise unit are included in the feasible degradation set. Units whose existence cannot be determined are saved as unresolved and are not merged into the proven set; as long as there are still unresolved units that may contain the allowed state of this group, the confidence level is reduced, the current state update is frozen, and a retest is requested, without issuing a tool correction permission based on the already solved local branches.

[0087] If none of the candidate degradation values ​​satisfy the load error bound, the feasible degradation set is set to empty. The feasible degradation set is also set to empty if the starting window or the recovery window is not marked as a valid window, no intermediate window is marked as a valid window, the chip thickness difference is below the identifiable threshold, or the window load difference between the starting window and the recovery window exceeds the repeatability error bound. The identifiable threshold is derived by inversely from the allowable uncertainty of the load difference error propagating to the material coefficients; excessively small thickness variations make the material coefficient estimates sensitive to noise, and updates are not permitted even if the equations are solvable. The repeatability error bound is determined by the measurement error bounds of the starting and recovery windows and the allowable short-term drift budget, and applies to load differences with the same thickness at the beginning and end, distinct from the load error bound used for single-window prediction residuals.

[0088] The consistency check only verifies whether the response can be repeated after restoring to the same cutting input. Local hard spots, inclusions, or cooling changes in intermediate positions may still pass the consistency check if they disappear before restoration; therefore, the main implementation method uses a defined homogeneous material and calibration process range, and simultaneously requires geometric intersection and model residual conditions for all windows. For workpieces with known local heterogeneity or those that cannot be covered by the error envelope, the effective life is not output; an intermediate window can be added to check its response, and the applicable range is still jointly limited by material conditions and model error envelope.

[0089] Only when the load difference between the first and last windows of the cutting window group meets the repeatability error bound and the feasible degradation set is not empty, is the degradation state and confidence level updated with the feasible degradation set, and the tool setting control signal output; otherwise, the degradation state is frozen and a retest is requested. The update record includes the tool lifecycle number, calibration package version, two measurement group numbers, original record index of each window, feasible degradation set, correlation between candidate and material coefficient intersection, and current cumulative cutting value. When frozen, the last valid record is retained, and the reason for this failure is also saved separately; failure observations are not used to rewrite the new tool reference or calibrate the normal response, nor do they trigger the retransmission of the previously permitted group.

[0090] The processing unit determines the reliability based on the span of the feasible degradation set, the proportion of the effective window, and the measurement timeliness of the cutting edge degradation interval. Reliability uses a rule level of effective, restricted, and invalid to represent the completeness of the current control basis. First, the probe, thermal calibration, and execution error budgets are deducted from the allowable tool position error. The remaining budget determines the upper limit of the feasible degradation set span; if the remaining budget is not positive, the reliability is set to invalid. The minimum effective proportion must ensure that different thicknesses and the start and recovery windows are all effective; this proportion is only satisfied when all three predetermined windows in the main implementation are effective. The maximum timeliness of geometric measurement is determined by dividing the remaining tool position error budget by the sum of the thermal displacement change rate boundary and the degradation rate boundary. If any rate boundary is unavailable or the sum of the two is not positive, no effective timeliness is published. It is effective when all three conditions are met. In addition to the listed invalid conditions, it is restricted when the set is not empty but any condition is insufficient; it is invalid when the set is empty, the model domain is out of bounds, or the measurement fails. When a feasible degenerate set has multiple interleaved branches, the determination is based on the outermost span and the status of unresolved branches, not the average branch value. This level does not represent statistical probability.

[0091] When the feasible degradation set is not empty, the feasible degradation set and the cumulative cutting time are stored in the degradation state. A degradation rate interval is formed by utilizing the change in candidate degradation amount with cumulative cutting time in multiple effective updates. The lower bound of the remaining life is determined based on the upper bound of the feasible degradation set, the upper bound of the degradation rate interval, and a preset tool change degradation amount threshold. The cumulative cutting time and calendar time are stored separately; the effective cutting time of the cutting edge in the same material and cutting domain is used to estimate the degradation rate. Each effective update saves the lower and upper ends of the feasible degradation set, while retaining the complete set. The rate range is formed using all allowed non-negative changes between multiple updates. Negative point estimates caused by noise are not interpreted as tool recovery. In the initial stage of a new tool, when there are not enough effective updates, controlled diagnostics are performed first, without predicting production life based on these diagnostics.

[0092] For two consecutive valid updates, first establish the allowable range of the average degradation rate based on the degradation range and the difference in cutting time:

[0093]

[0094] Among them, the allowable range of average degradation rate Measured in millimeters per minute; the lower end of this degradation and the top The lower end of the previous degeneration and the top All measurements are in millimeters; effective cumulative cutting time difference Measured in minutes and must be greater than 0, take the larger value. This is used to limit non-negative backsliding. If the entire range is lower than the previous range and cannot be explained by error, it is first judged as inconsistent and rejected for rate formation; this set of data is not included in the above non-negative rate calculation. Adjacent intervals provide input for the subsequent predicted rate envelope and are not directly equivalent to future instantaneous rates.

[0095] The observation envelope is formed by the allowable rate ranges of the most recent adjacent updates, and the rate variation margin in the calibration under the same operating conditions and the predetermined prediction time domain is added to obtain the degradation rate interval. The upper bound of the prediction is not less than the sum of the upper bounds of the average rates used and the variation margin; the variation margin is determined based on the accelerated degradation range in the independent diagnostic records and cannot be zero simply because the fitted curve is smooth. When the next task exceeds the calibrated prediction time domain, material range, or cutting parameter domain, the degradation rate interval is no longer used for life prediction.

[0096] Given an effective positive rate upper bound and the next task falling within the applicable operating conditions and prediction time domain, calculate the lower bound of the remaining cutting time before reaching the tool change degradation threshold:

[0097]

[0098] Among them, the lower bound of remaining lifetime In minutes; preset tool change degradation threshold and the upper bound of the current feasible degenerate set In millimeters; the upper bound of the degradation rate range Measured in millimeters per minute, take the larger value. The result when the threshold is reached is set to 0. The tool change threshold is determined by the allowable dimensional deviation, tool edge usage limitations, and the calibrated retraction range. This lower bound is used to determine whether the next task can be completed, provided that the rate envelope continues to cover subsequent working conditions; it is not used to guarantee that sudden tool breakage will not occur.

[0099] When the feasible degradation set is empty, the previously valid degradation state is maintained. When the feasible degradation set is empty, the upper bound of the degradation rate interval cannot be determined, or the machining conditions exceed the applicable range of the degradation rate estimation, the confidence level is reduced and a lifetime unavailable state is output. When the feasible degradation set is not empty and the current geometry and load update conditions are met, the current degradation record is saved. If the rate has not yet been established, lifetime prediction is made unavailable, and subsequent valid records obtained by controlled diagnosis are still used to form the degradation rate interval. Cutting time that continues to occur after valid measurement is deducted from the available lifetime and included in the measurement aging; freezing the geometry degradation state does not freeze the cumulative cutting time. When the historical record expires, the upper bound of the rate is zero, the time difference is zero, or the speed domain changes, the remaining lifetime must not be set to infinity. When lifetime is unavailable, production feed is suspended, and a preset retest and diagnostic process is allowed to be executed under the interlock control of the spindle, probe, and tool holder. Production permission is restored after a sufficiently valid state is formed.

[0100] The tool setting control signals include offset correction commands, spare tool switching commands, or stop and retest commands. Stop and retest commands take precedence over spare tool switching commands, and spare tool switching commands take precedence over offset correction commands. The processing unit prioritizes stop conditions, then tool change conditions, and finally offset correction conditions; only one command is issued per cycle. Each command carries the tool edge identifier, measurement group number, target offset version or spare tool position, validity period, and sequence number. The CNC system executes the command after confirming the current tool edge and machining status, and sends back success, failure, or timeout status and the effective time; duplicate sequence numbers are not executed repeatedly.

[0101] When the reliability meets the preset conditions and the upper bound of the feasible degradation set is less than the preset tool change degradation threshold, the processing unit generates the limited offset correction command based on the thermal drift measured by the tool setting and the online measurement deviation. The online measurement deviation is given by the workpiece online measurement probe or the on-machine measurement interface connected to the CNC system, representing the signed difference between the machined dimension and the target dimension, and carrying the measurement direction and time. The tool setting thermal drift comes from the common drift of the reference surface and the differential elongation, and is converted into the displacement in the machining radius direction through the calibration mapping of the machine tool measurement direction, the workpiece clamping datum, and the controlled dimension; the tool holder reference surface drift cannot directly represent the total thermal error between the workpiece and the tool tip.

[0102] To avoid compensating for the thermal effects already included in the online dimensional deviations, only the thermal displacement added after the dimensional measurement is combined with the measured dimensional deviation to form a correction amount:

[0103]

[0104] Among them, physical radial offset increment Non-negative single-off bias limiting In millimeters; online dimensional tolerances Subtract the target size from the actual size, in millimeters; calibrated sensitivity of the size to radial offset. Dimensionless and non-zero; the additional equivalent thermal displacement after dimensional measurement. In millimeters, its positive direction is defined as the direction in which a negative radial offset needs to be canceled. (Minimum and maximum calculations) , A symmetrical limiting is established. When measuring the outer diameter and the positive physical radial offset increases the radius, the sensitivity is 2; for other directions, values ​​are taken according to the calibration sign. When the CNC interface uses diameter tool compensation, the physical radial offset increment is converted to the corresponding diameter increment before writing.

[0105] When the lower bound of the remaining lifespan is less than the estimated cutting time of the next machining task, or when the upper bound of the feasible degradation set reaches the preset tool change degradation threshold, a spare tool switching command is generated, pointing to a spare tool that has passed tool setting verification. The spare tool first completes clamping identification, probe tool setting, and new tool reference association; before switching, it saves the current cutting edge state; after the feed stops, it calls the spare tool position and waits for turret locking or tool positioning confirmation. If there is no spare tool that has passed tool setting verification, the tool change condition changes to a stop and retest. The spare tool uses its own lifespan record and process calibration package and cannot carry over the cumulative cutting time of the replaced cutting edge.

[0106] When the reliability does not meet the preset conditions, the lifespan becomes unavailable, the online measurement deviation still exceeds the limit after offset correction, or the spare tool verification fails, the stop and retest command is generated. After offset correction, the workpiece dimensions are retested at the next allowed in-machine measurement position. If the deviation still exceeds the limit, the offset and measurement conditions before and after correction are recorded, and the continued accumulation of offset is stopped. Clamping changes, probe contamination, workpiece thermal deformation, and model mismatch are investigated first. A new permission is issued only when the retest is successful and the status is valid. The stop state is maintained if the measurement is missing or the execution confirmation times out. The single amplitude limit is determined based on the workpiece allowance, the maximum allowable size change, and the tool setting repeatability. The cumulative correction limit cannot be bypassed by multiple fast writes.

[0107] The processing unit associates and stores the offset values ​​before and after offset correction, as well as the effective time, with the measurement coordinates. Offset correction does not reset the degradation state. After blade replacement, a new blade reference and degradation state are established. The new blade reference, calibration package, and degradation state are each labeled with a version number. The measurement record saves the actual effective version at the time of measurement, not the latest blade compensation value at the time of query. Blade replacement only establishes a reference for the new blade's lifecycle; the old history is used for tracing. After ordinary offset correction, the original coordinates are restored, and the resulting physical degradation relationship should remain consistent within the measurement error range.

[0108] The signal acquisition component associates the cutting window group and window load of the cutting edge on the same workpiece according to the program segment. The chip thickness of the cutting window group changes first and then recovers. In an independently deployed online tool setting device, the signal acquisition component reads the actual magnification, spindle load, and cutting edge identification within the same program segment from the CNC system. The processor binds the window record with the preceding and following measurement group numbers. Tool setting correction permission is not issued for spindle and auxiliary spindle load attribution, measuring arm position, or clamping status that the device itself cannot confirm.

[0109] The processor controls the online tool setting probe to sequentially measure the non-cutting reference surface, the cutting edge, and the non-cutting reference surface before and after the cutting window group, restoring the measured coordinates to the same offset reference. It then subtracts the interpolation drift of the measured coordinates of the non-cutting reference surface and the differential elongation calibrated according to the temperature sensor data, forming a corresponding post-group cutting edge degradation interval relative to the new tool reference. The processor can latch the coordinates and offset in the probe-triggered interrupt and call the geometric calculation program after the measurement sequence is completed. This program returns the interpolation, differential elongation, and error propagation results as the pre-group or post-group cutting edge degradation interval. The same new tool reference is shared at both time points, and the identity of the temperature calibration package is bound to the tool holder, tool extension length, and measurement direction.

[0110] The processor selects candidate degradation values ​​within the intersection of the degradation intervals of the cutting edge. It uses a calibration load model that simultaneously incorporates material coefficients for both the shearing and cutting edge terms, retaining candidate degradation values ​​that can explain the load of each window within the load error bound using the same material coefficient, thus forming a feasible degradation set. The device employs the aforementioned continuous interval solution or piecewise bounded solution, writing the common material coefficient set and failure reason for each candidate to local storage. When the load difference between the first and last windows of the cutting window group satisfies the repeatability error bound and the feasible degradation set is not empty, a tool setting correction permission is output according to the feasible degradation set; otherwise, a retest is requested. The tool setting correction permission is only the output that is compatible with the model for this group of measurements; when connected to the system, the processing unit must also determine whether to actually write the offset based on reliability, lifetime, and control priority. The permission must be consistent with the current cutting edge and offset version to prevent old permissions from acting on new machining states.

[0111] To illustrate the material conditional calculation, the equivalent cutting edge length of the new blade is taken as 0.02 mm, the material conditional degradation coefficient as 0.5, the additional degradation response coefficient as 200 N / mm², the initial and recovery window thicknesses as 0.1 mm, the intermediate window thickness as 0.2 mm, and the blade degradation as 0.02 mm. When the material coefficients are 800, 1000, and 1200 N / mm², the initial and intermediate window loads are 108 and 188, 134 and 234, and 160 and 280 N / mm², respectively; the recovery window is taken as the corresponding load under the same thickness. The material coefficients can be restored by dividing the load difference between the two thicknesses by the thickness difference, and then substituting them into the cutting edge relationship, all yielding a degradation of 0.02 mm. Material changes simultaneously alter both shear and cutting edge loads; under different material coefficients, the set degradation amount is restored according to their respective shear and cutting edge terms.

[0112] In another set of input settings with error bounds, the aforementioned cutting edge model is used, with the thicknesses of the three windows being 0.1, 0.2, and 0.1 mm, and the loads being 134, 234, and 134.1 N / mm, respectively. The load error bound for each window is 0.2 N / mm, the material coefficient range is 800 to 1200 N / mm², and the geometric candidate range is 0.019 to 0.022 mm. After simultaneously setting the material coefficient ranges, the lower end of the feasible degenerate set is rounded out to 0.019301496 mm, and the upper end is rounded out to 0.021031519 mm. The rounded endpoints are used to store the numerical representation range, and the existence of candidates is determined by the unrounded constraints. When the candidate degenerate amount is 0.020 mm, the common material coefficient range is approximately 999.230769 to 1000.869566 N / mm². If the repeatability error limit is 0.5 N / mm, the first-to-last difference of 0.1 N / mm passes the repeatability test.

[0113] If the recovery window load is changed to 136 N / mm, and the difference between the first and last measurements is 2 N / mm, exceeding the repeatability error limit of 0.5 N / mm, then the feasible degenerate set is set to empty and a retest is requested, even if the initial and intermediate windows can still be algebraically solved, the state is not updated. If the allowable ranges for the geometric measurements are 0.018 to 0.022 mm and 0.025 to 0.029 mm respectively, and the geometric intersection is empty, then no update is performed. If the two thicknesses are the same, any critical measurement moment is repeated, the probe trigger sequence is incomplete, or the tool compensation version changes during the measurement, the measurement is rejected at the identifiability, temporal sequence, or group integrity checks, respectively.

[0114] Using another set of valid historical data to illustrate lifespan and control linkage: Under the same operating conditions, the previous degradation range was 0.014 to 0.016 mm. After 10 minutes of effective cutting, the current range is 0.019 to 0.021 mm. Therefore, the allowable range of adjacent average degradation rates is 0.0003 to 0.0007 mm per minute. Assuming that the remaining valid updates and the calibrated rate change margin make the upper limit of the predicted rate 0.0008 mm per minute, and the tool change degradation threshold is 0.050 mm, then the lower limit of the remaining lifespan is 36.25 minutes. When the next task is expected to cut for 40 minutes, if there are no higher priority shutdown conditions, switch to the standby tool that has been verified. If the predicted upper limit cannot be determined to be positive or the material for the next task exceeds the calibration package, the output lifespan is unusable, and the machine is stopped for retesting, without using the 36.25-minute result.

[0115] A complete machining cycle begins with a check of the current cutting edge identity and calibration package validity: the tool retracts to the measurement position, the pre-measurement group is completed, and the probe is retracted; within the diagnostic allowance of the same workpiece, the cutting window group is restored after the thickness change; the spindle is stopped again, and the pre-measurement group is completed. When the geometric intervals, window loads, and temperature records are compatible, a feasible degradation set and reliability are formed, and the lifespan is updated. Assuming the online dimensional deviation of the outer diameter is 0.012 mm, the added equivalent thermal displacement after measurement is 0.001 mm, the dimensional sensitivity to radial offset is 2, and the single-shot limit is 0.010 mm, then the physical radial correction is −0.007 mm; using the diameter tool compensation interface, −0.014 mm is written. After execution confirmation, the time when the new offset takes effect is recorded. Subsequent machining only resumes when the workpiece remeasurement is qualified and the probe exits and is confirmed to be valid.

[0116] For horizontal turret lathes, the tool holder rotates with the turret; turret locking and tool position confirmation are read before measurement. For multi-tool lathes, a unidirectional approach path is planned based on the fixed position of the tool holder. When a Swiss-type lathe uses a guide sleeve, the calibration package includes guide sleeve constraints and bar stock extension status; after switching to guide sleeve-free mode, the corresponding calibration package is used and the same-stage window relationship is re-established. The processing unit can be located within the CNC system or within a computing unit connected via an industrial communication interface. When externally located, the spindle stop, probe position, and feed prohibition interlocks are retained on the machine tool side; communication interruption invalidates unexecuted permissions. All the above arrangements use the same cutting edge's front and rear geometric constraints, common material coefficients, and recovery window gating, without changing the separation relationship between the degradation state and the offset history.

Claims

1. A decoupled monitoring system for CNC lathe tool degradation and thermal drift, characterized in that, The system includes a tool setting unit, a data acquisition unit, and a processing unit connected to the CNC system. The data acquisition unit associates cutting window groups and window loads on the same workpiece and the same cutting edge according to program segments. The chip thickness of the cutting window group is first changed and then restored. The tool setting unit sequentially measures the non-cutting reference surface on the tool holder carrying the cutting edge, the cutting edge, and the non-cutting reference surface before and after the cutting window group, respectively. It restores the measured coordinates to the same offset reference, deducts the interpolation drift of the measured coordinates of the non-cutting reference surface and the differential elongation due to temperature calibration, and forms a reference relative to the new tool reference. The processing unit selects candidate degradation amounts within the intersection of the degradation ranges of the cutting window group, uses the material coefficient to simultaneously participate in the calibration load model of the shear term and the cutting edge term, and retains the candidate degradation amounts that can explain the load of each window within the load error bound with the same material coefficient, forming a feasible degradation set; only when the load difference between the first and last windows of the cutting window group satisfies the repeatability error bound and the feasible degradation set is not empty, the degradation state and confidence are updated with the feasible degradation set and the tool setting control signal is output; otherwise, the degradation state is frozen and a retest is requested.

2. The system according to claim 1, characterized in that, The acquisition unit uses program version, program segment number, tool edge identifier, spindle channel, and clamping status to form a stage identifier. According to the stage identifier, cutting time, spindle load, servo current, vibration, temperature, and online measurement deviation are mapped to the normalized cutting process within the program segment. Each window in the cutting window group has the same stage identifier. For Swiss-type lathes, the guide sleeve mode is included in the stage identifier, and the main and auxiliary spindle handover, tool change, and parallel cutting intervals that cannot be separately attributed to tool edge loads are excluded.

3. The system according to claim 1, characterized in that, The non-cutting reference surface and the cutting edge are located on the same locked tool holder. The tool setting unit records the coordinates, time, and effective offset value of each measurement group under the same measurement direction and trigger speed. The measurement coordinates are restored using the effective offset value. The non-cutting reference surface position of the cutting edge at the measurement time is obtained by interpolation according to the time ratio between the measurement time of the cutting edge and the measurement time of the two non-cutting reference surfaces. After the displacement change of the cutting edge relative to the position of the non-cutting reference surface is corrected by the differential elongation, the cutting edge position degradation amount relative to the new tool reference is determined along the cutting edge retraction direction. The probe repeatability error, interpolation error, offset restoration error, differential elongation calibration error, new tool reference error, and intra-group retraction change error are propagated into the cutting edge position degradation interval.

4. The system according to claim 3, characterized in that, The differential elongation is the thermal elongation of the cutting edge relative to the non-cutting reference surface. It is calibrated by the displacement change of the cutting edge relative to the non-cutting reference surface, the tool holder temperature, and the near-cutting temperature of the tool during heating and cooling without cutting. The heating and cooling processes are respectively used to establish calibration relationships. The processing unit selects the calibration relationship according to the direction of temperature change. If the temperature exceeds the calibration range, the direction of temperature change cannot be determined, or the difference between the coordinates of two measurements of the non-cutting reference surface exceeds the allowable drift, the processing unit refuses to let the cutting edge degradation interval participate in the update and requests a retest.

5. The system according to claim 1, characterized in that, The cutting window group includes a starting window, at least one intermediate window, and a recovery window arranged in sequence. The starting window, the intermediate window, and the recovery window are located in adjacent equal depth of cut regions of the same workpiece, and have the same cutting speed, cutting width, and cooling state. The starting window and the recovery window use the same undeformed chip thickness, while the intermediate window uses a different undeformed chip thickness than the starting window. The processing unit deducts the no-load load at the same rotation speed, converts the spindle load into a tangential force per unit cutting width to constitute the window load, and uses servo current and vibration to mark windows that pass steady-state detection as valid windows. When the starting window or the recovery window is not marked as a valid window, no intermediate window is marked as a valid window, the chip thickness difference is lower than the identifiable threshold, or the window load difference between the starting window and the recovery window exceeds the repeatability error limit, the feasible degradation set is set to empty.

6. The system according to claim 5, characterized in that, The calibration load model takes the candidate degradation amount, the material coefficient, and the undeformed chip thickness as inputs. The shear term is the product of the material coefficient and the undeformed chip thickness, and the cutting edge term is the calibration response that varies with both the candidate degradation amount and the material coefficient. The shear term and the cutting edge term are added together to obtain the predicted window load. For each candidate degradation amount, the processing unit calculates, within the material coefficient calibration range, the set of material coefficients that ensures the difference between the predicted window load and each window load falls within the load error bound formed by measurement error, model calibration error, and intra-group shrinkage propagation error. Candidate degradation amounts whose intersection with each set of material coefficients is not empty are included in the feasible degradation set, and the correlation between each candidate degradation amount in the feasible degradation set and the intersection with the material coefficient is saved. When all candidate degradation amounts do not satisfy the load error bound, the feasible degradation set is set to empty.

7. The system according to claim 6, characterized in that, The processing unit determines the reliability based on the span of the feasible degradation set, the proportion of the effective window, and the measurement aging of the cutting edge degradation interval. When the feasible degradation set is not empty, the feasible degradation set and the cumulative cutting time are stored in the degradation state. A degradation rate interval is formed by utilizing the change of candidate degradation amount with cumulative cutting time in multiple effective updates. The lower bound of the remaining life is determined based on the upper bound of the feasible degradation set, the upper bound of the degradation rate interval, and a preset tool change degradation amount threshold. When the feasible degradation set is empty, the previously effective degradation state is maintained. When the feasible degradation set is empty, the upper bound of the degradation rate interval cannot be determined, or the machining conditions exceed the applicable range of the degradation rate estimation, the reliability is reduced and a life unavailable state is output.

8. The system according to claim 7, characterized in that, The tool setting control signal includes an offset correction command, a spare tool switching command, or a stop and retest command. The stop and retest command takes precedence over the spare tool switching command, and the spare tool switching command takes precedence over the offset correction command. When the confidence level meets the preset conditions and the upper bound of the feasible degradation set is less than the preset tool change degradation threshold, the processing unit generates the offset correction command with a limit based on the thermal drift measured by the tool setting and the online measurement deviation. When the lower bound of the remaining lifespan is less than the expected cutting time of the next machining task or the upper bound of the feasible degradation set reaches the preset tool change degradation threshold, the spare tool switching command pointing to the spare tool that has passed the tool setting verification is generated. When the confidence level does not meet the preset conditions, the lifespan becomes unavailable, the online measurement deviation still exceeds the limit after offset correction, or the spare tool verification fails, the stop and retest command is generated. The processing unit stores the offset values ​​before and after offset correction and the effective time in association with the measurement coordinates. Offset correction does not reset the degradation state. After the tool is replaced, a new tool reference and degradation state are established.

9. An online tool setting device, characterized in that, The system includes an online tool setting probe, a tool holder equipped with a cutting edge and a non-cutting reference surface, a temperature sensor, a signal acquisition component, and a processor connected to a CNC system. The signal acquisition component associates the cutting window group and window load of the cutting edge on the same workpiece according to program segments. The chip thickness of the cutting window group is first changed and then restored. The processor controls the online tool setting probe to sequentially measure the non-cutting reference surface, the cutting edge, and the non-cutting reference surface before and after the cutting window group, restoring the measured coordinates to the same offset reference and subtracting the interpolation of the measured coordinates of the non-cutting reference surface. The drift and differential elongation calibrated according to the temperature sensor data form the corresponding post-group state of the cutting edge degradation interval relative to the new tool reference; the processor selects candidate degradation amounts within the intersection of the cutting edge degradation intervals, and uses the material coefficient to simultaneously participate in the calibration load model of the shear term and the cutting edge term, retaining the candidate degradation amounts that can explain the load of each window within the load error boundary with the same material coefficient, forming a feasible degradation set; when the load difference between the first and last windows of the cutting window group satisfies the repeatability error boundary and the feasible degradation set is not empty, the tool setting correction permission is output according to the feasible degradation set; otherwise, a retest is requested.

10. The online tool setting device according to claim 9, characterized in that, It also includes a vibration sensor fixed to the tool holder, and temperature sensors are respectively arranged on the tool holder and the near-cutting edge of the tool. The signal acquisition component identifies the unsteady window in the cutting window group according to the signal of the vibration sensor. When an unsteady window exists, the processor sets the feasible degenerate set to empty. The processor only allows group measurement when the spindle stops, the tool holder is locked, and the online tool setting probe is in position signal is valid. If the effective bias value changes or the measurement sequence is incomplete during the group measurement, the group measurement is discarded, and the machining feed prohibition state is lifted after the confirmation signal of the online tool setting probe exiting the measurement position is valid.

Citation Information

Patent Citations

  • Method for determining wear state of rear cutter surface of cutter based on variable-feed trial cutting

    CN113601264A

  • Mechanism-data fusion driven variable working condition tool wear state monitoring method

    CN114102260A

  • Measuring method and device, machine tool having such device, and work processing method

    EP1327495A1

  • Production tool wear detector

    US4120196A