A machine tool feeding system friction deterioration positioning method based on spatial energy consumption hysteresis loop and sliding window AUC integral

Abstract

This invention discloses a method for locating friction degradation in machine tool feed systems based on spatial energy consumption hysteresis loops and sliding window AUC integration, belonging to the field of CNC machine tool fault diagnosis. The invention includes: acquiring multi-source data and performing sub-millimeter-level synchronization; extracting low-speed crawling steady-state conditions; decoupling the current signal in time and space and mapping it to a unified spatial grid; constructing forward and reverse spatial energy consumption hysteresis loops and sub-segment loops; calculating the absolute value integral of the forward and reverse current difference, AUC(s), using a sliding window; and outputting the spatial coordinates of friction degradation when AUC(s) exceeds the dynamic threshold and the hysteresis loop expands / changes at high frequencies. This invention completely decouples the time axis, eliminates interference from acceleration and deceleration inertial currents, achieves sub-millimeter-level positioning (error ±1mm), and can accurately identify the "stick-slip" phenomenon and lubricating oil film rupture in the low-speed range. It solves the problems of fuzzy positioning and insensitivity to low-speed friction in traditional time-domain diagnostics, and is suitable for intelligent operation and maintenance of CNC machine tool feed systems.

Description

Technical Field

[0001] This invention belongs to the fields of dynamic performance monitoring of CNC machine tools, parameter optimization of high-end CNC systems, and intelligent operation and maintenance technology in industrial sites. Specifically, it relates to a method for reconstructing hysteresis loops in the spatial domain using servo current data and combining it with the integration of the absolute value of the sliding window area to achieve sub-millimeter-level precise positioning of machine tool friction degradation. Background Technology

[0002] The linear feed system of CNC machine tools (mainly composed of servo motors, ball screws, linear guides, sliders, worktables, etc.) generally suffers from the following steady-state and quasi-static friction degradation problems during long-term high-intensity reciprocating machining:

[0003] Non-parallel installation of mechanical guideways can cause interference in specific locations of the slider, localized wear at the limit stroke leading to pits or carbon deposits, and rupture of the lubricating oil film resulting in boundary friction or even dry friction. When these deteriorations develop to a certain extent, they can cause motion stagnation or low-speed crawling vibration (i.e., stick-slip) at specific spatial locations, directly affecting the accuracy of the machined surface morphology, and in severe cases, causing workpiece scrap or tool damage.

[0004] Traditional machine tool diagnostic methods typically anchor control current and vibration signals on the "time axis," which faces the following fundamental engineering bottlenecks:

[0005] In actual testing or machining, the transient inertial disturbances during acceleration and deceleration completely overwhelm the frictional characteristics. Feed rates frequently fluctuate dramatically between zero speed, low-speed crawling (e.g., F10, 10 mm / min), rapid displacement (e.g., F3000, 3000 mm / min), and high-speed positioning (e.g., F10000, 10000 mm / min). The current signal on the time axis generates a massive transient control load during acceleration and deceleration. This load primarily overcomes the inertial torque, and its amplitude is often several times, or even an order of magnitude, greater than the current change caused by steady-state friction. For example, during an emergency stop at F10000, the current spike can reach over 5A, while the current change caused by low-speed steady-state friction is typically only 0.2~0.5 A. Therefore, directly analyzing the current signal on the time axis completely overwhelms the weak steady-state resistance caused by uneven local friction distribution, making it impossible to effectively extract.

[0006] The spatial positioning capability is severely lacking. Traditional methods record signals on a time axis. When abnormal current or vibration characteristics are detected, the unstable machine tool speed (speed variations cause the same physical displacement to correspond to different time spans) makes it difficult to accurately map the extracted time feature points back to the machine tool's actual physical coordinates. For example, a 100-millisecond current fluctuation corresponds to a physical travel of only about 0.017 mm at a feed rate of 10 mm / min, but about 16.7 mm at a feed rate of 10000 mm / min. In actual speed-changing tests, different speed ranges result in waveforms of the same width on the time axis potentially corresponding to travel differences of tens or even hundreds of millimeters. This leads to extremely poor positioning accuracy, making it impossible to guide on-site maintenance personnel in targeted local troubleshooting and repair (such as local scraping, unclogging specific oil passages, or replacing local protective covers).

[0007] Vibration signals are completely insensitive to frictional forces at low speeds. Under low-speed crawling conditions (such as F10), due to the extremely low speed and minimal impact energy, the vibration amplitude collected by the accelerometer is typically less than 0.05 g (gravitational acceleration), and there are almost no positional fluctuations related to changes in frictional force throughout the entire stroke. Relying solely on vibration analysis would lead to the erroneous conclusion of "no abnormality," resulting in missed diagnoses.

[0008] To address the aforementioned shortcomings, this invention proposes a diagnostic method that completely decouples the time axis and is based on a spatial energy consumption hysteresis loop and a sliding window AUC integral. The core idea of ​​this method is to specifically design a low-speed steady-state reciprocating crawling condition so that the current signal only reflects steady-state frictional force; then, to completely discard the time variable and bind the current to the position; by constructing a spatial hysteresis loop between forward and reverse currents and utilizing a sliding window integral to amplify the slight difference in frictional energy consumption into a significant AUC step, thereby achieving sub-millimeter-level precise positioning of abnormal frictional force distribution throughout the entire stroke. Summary of the Invention

[0009] The purpose of this invention is to overcome the blindness and ambiguity of time-domain diagnostics and provide a machine tool friction degradation positioning method based on spatial energy consumption hysteresis loop and sliding window AUC integration. By designing a dedicated low-speed steady-state crawling excitation condition, the time independent variable is completely decoupled, and the forward and reverse feed currents are mapped to a unified standard spatial physical grid. A "full-stroke spatial energy consumption hysteresis loop" is constructed to characterize the overcoming of static friction. Using the area absolute value integral model of the sliding window, the weak friction degradation is transformed into an abnormal step curve, thereby achieving automated, sub-millimeter-level precise spatial positioning of non-uniform friction degradation throughout the entire stroke range.

[0010] A method for locating friction degradation in a machine tool feed system based on spatial energy consumption hysteresis loop and sliding window AUC integration, characterized by the following steps:

[0011] S1, Physically Isolated Data Synchronization and Working Condition Segmentation

[0012] The external high-frequency vibration signal and internal servo control parameters are collected under multiple working conditions of the machine tool. The internal servo control parameters include at least the real current signal IEFF, the speed command VCMD, and the feedback position POSF.

[0013] S1a. Physical isolation measures: When collecting external high-frequency vibration signals, an electrical insulating medium (such as standard A4 paper, 0.05~0.15 mm thick) must be sandwiched between the sensor adsorption base and the metal surface of the machine tool to cut off the conductive interference introduced by the high-frequency chopping of the machine tool; both the external acquisition equipment (data acquisition instrument, laptop) and the internal monitoring equipment are powered by independent DC power modules to eliminate random electromagnetic crosstalk introduced by the power grid.

[0014] S1b, Impact Synchronization Anchor Point: In the excitation program, a combination of G04 pause and G01 shift commands is inserted at the start and end points to artificially generate a physical impact waveform. The point where the maximum value of the first derivative of the servo torque command is extracted is used as the electrical signal anchor point T_elec, and the rising edge intersection point where the triaxial composite vibration energy exceeds a preset threshold for background noise (e.g., 5 times the mean of background noise) is extracted is used as the mechanical anchor point T_mech. The delay offset between the two is calculated using a cross-correlation function.

[0015]

[0016] The vibration sequence is shifted forward by Δt to compensate. Then, a one-dimensional cubic spline interpolation algorithm is used to upsample and resample the intrinsic servo parameters (IEFF, VCMD, POSF) to make their sampling rate consistent with the vibration signal (e.g., 10240 Hz), thereby achieving sub-millimeter-level precise synchronization of multi-source heterogeneous signals at the same frequency reference.

[0017] S1c, Working Condition Semantic Segmentation: Constructing a finite state machine based on the absolute value of the speed instruction VCMD(t).

[0018] When |VCMD| < 2 mm / min, it is determined to be Idle (stationary);

[0019] When 2 ≤ |VCMD| < 5 mm / min, it is determined to be Moving (low-speed transition);

[0020] When 5 ≤ |VCMD| < 50 mm / min and the duration is ≥0.1 seconds, it is judged as LowSpeed ​​(low-speed crawling steady-state condition).

[0021] When |VCMD| ≥ 9000 mm / min, it is determined to be HighSpeed. All LowSpeed ​​semantic blocks are extracted as the sole data source for subsequent analysis.

[0022] S2, Time-Space Axis Decoupling and Unified Spatial Grid Resampling

[0023] Complete decoupling of time-space axes and reconstruction of spatial geometric hysteresis loops: Completely eliminate the time independent variable t, extract semantic block data of low-speed crawling conditions, and establish a spatial transformation mapping from time-domain sampling points to actual feedback positions. Projecting the actual current amplitudes (IEFF) of the forward crawling segment and the reverse reciprocating crawling segment onto the spatial axis, a bidirectional spatial energy dissipation hysteresis loop geometry of the entire stroke (current-position) is constructed:

[0024]

[0025] Where s is the discrete travel coordinate of the spatial axis. For a positive current trajectory, This is the reverse current trajectory; after this step, the forward and reverse currents are spatially aligned point by point and can be directly compared.

[0026] S3. Construction of full-journey spatial energy consumption hysteresis loop and decoupling of sub-segments

[0027] Using the same spatial coordinates s (s) and (s) are plotted in the same coordinate system with POSF as the abscissa and IEFF as the ordinate, forming a "full-stroke bidirectional spatial energy consumption hysteresis loop".

[0028] The physical meaning of this hysteresis loop is: at any position s, the difference between the frictional force required to overcome for forward motion and the frictional force required to overcome for reverse motion directly reflects the asymmetry of bidirectional resistance at that position. The longitudinal span of the loop (i.e., (s) and The larger the distance between (s), the higher the additional servo energy required to overcome friction at that location.

[0029] Furthermore, based on the machine tool's physical feed clearance (for example, five independent reciprocating segments were designed for the X-axis low-speed crawling test, corresponding to 5~90 ​​mm, 90~180 mm, 180~270 mm, 270~360 mm, and 360~437 mm respectively), the full-stroke hysteresis loop is automatically cut and decoupled into several independent and continuous local sub-segment hysteresis loops. Each sub-segment corresponds to one complete forward + reverse reciprocating motion.

[0030] This decoupling can eliminate the influence of current baseline drift caused by different starting points between different sections, making it easier to independently assess the friction state of each section.

[0031] S4. Integral of the absolute value of the spatial energy consumption envelope area (AUC) based on the sliding window.

[0032] Set a spatial sliding window with a length of W and a step size of Δd.

[0033] In this invention, W is preferably 50 mm (approximately 2 to 5 times the lead screw pitch, which can smooth local noise while preserving spatial resolution), and Δd is preferably 10 mm (to ensure the sensitivity of abnormal interval edge detection).

[0034] Within each sliding window [s - W / 2, s + W / 2], the absolute area AUC(s) enclosed by the forward and reverse current curves is calculated using the trapezoidal integral method. Its physical meaning is the net servo energy (unit: A·mm) consumed by the feed system to overcome steady-state friction within this local spatial segment. The calculation formula is as follows:

[0035]

[0036] Because of I (x) and All values ​​are positive real numbers (absolute values ​​already taken), and the forward current is usually greater than the reverse current (to overcome the resistance of the guide rail), so the integrand is positive and the integral result is positive. This integral value should remain basically constant for the healthy axis throughout its entire stroke; when there is frictional degradation in a certain spatial interval, the difference between the forward and reverse currents increases significantly, causing AUC(s) to show local peaks or continuously rise.

[0037] S5, Spatial Anomaly Detection and Millimeter-Level Degradation Location

[0038] By comparing the spatial energy area distribution curve AUC(s), the mean AUC μ and standard deviation σ of the stationary baseline segment are extracted as the health benchmark. The selection principle for the stationary baseline segment is to use AUC data from the middle section of the entire stroke (e.g., 150~300 mm on the X-axis). This area is typically the frequently used range for daily machine tool processing, with relatively uniform wear, and can serve as a health reference. The dynamic envelope threshold is set to μ + k・σ, where k is a preset coefficient, preferably k=3 (corresponding to a 99.7% confidence interval; values ​​exceeding this range are considered statistically significant anomalies).

[0039] When the AUC(s) of a specific physical space interval continuously (for at least 3 consecutive windows) exceeds the dynamic envelope threshold, and the corresponding local sub-segment hysteresis loop exhibits significant expansion in its longitudinal span (loop width increases by more than 50% compared to the healthy segment) or high-frequency jumps (irregular sawtooth patterns with amplitude > 0.05A and frequency > 10Hz appear in the current waveform), the system automatically determines that there is guide rail friction degradation or lubricating oil film rupture in that space interval. The final output is a diagnostic physical coordinate, for example: "Severe friction degradation exists in the X-axis range of 375 mm ~ 425 mm; it is recommended to check the guide rail surface and oil passages in this range."

[0040] Technical effectiveness verification:

[0041] Completely eliminate acceleration and deceleration inertial interference: By using only low-speed steady-state crawling conditions (F10, i.e., 10 mm / min) and completely eliminating the time axis, the current signal corresponds entirely to the steady-state friction force, and the transient components of acceleration and deceleration are excluded from the analysis.

[0042] Sub-millimeter level spatial positioning capability: Sliding window integration locks abnormal locations within half the window width (±25 mm). Combined with spatial grid interpolation (1 mm resolution), the actual positioning error is better than ±1 mm, which is far superior to traditional time-domain methods (usually tens of millimeters of error).

[0043] Quantitative diagnostic indicator: The integral value of AUC(s) directly corresponds to the energy consumption of the local space. It can automatically alarm through the statistical threshold (μ+3σ) without relying on human experience for interpretation, making it suitable for embedding in automated operation and maintenance systems.

[0044] Microscopic friction condition identification: By identifying the high-frequency burr characteristics of the hysteresis loop in a local sub-segment, the "stick-slip" phenomenon can be identified, determining whether the lubricating oil film is broken, and thus distinguishing between different fault modes such as "dry friction" and "hard interference".

[0045] Strong resistance to electromagnetic interference: Through physical isolation (A4 insulating paper, independent battery power supply) and sub-millimeter-level cross-correlation synchronization algorithm, it effectively suppresses electromagnetic crosstalk and clock asynchrony in the factory environment.

[0046] High repeatability and robustness: The hysteresis loop morphology and AUC curve trend of the three repeated experiments (Rep1, Rep2, Rep3) are highly consistent, proving that the method is not affected by random noise and the diagnostic conclusions are stable. Attached Figure Description

[0047] Figure 1 This is a flowchart of a method provided in an embodiment of the present invention.

[0048] Figure 2 This is a schematic diagram of the hysteresis loop of the full stroke of the feed axis (Y-axis) of a healthy machine tool provided in an embodiment of the present invention; wherein, the solid line is the positive real current and the dashed line is the reverse real current; the horizontal axis is the actual feedback position POSF (mm) and the vertical axis is the real current IEFF (A).

[0049] Figure 3 is a schematic diagram of the spatial energy consumption hysteresis loop of the entire stroke of the feed axis (X-axis) of the friction-degraded machine tool provided in an embodiment of the present invention, showing the spatial characteristics of the divergent expansion of the hysteresis loop in the right limit stroke region.

[0050] Figure 4 is a schematic diagram comparing the hysteresis loops of the friction-degraded feed shaft provided in the embodiment of the present invention after it is divided into multiple local sub-segments, showing the high-frequency stick-slip waveform characteristics of the severely degraded region (segments 4 and 5).

[0051] Figure 5 is a schematic diagram of the energy consumption envelope area AUC(s) curve obtained by the absolute area integration of the spatial sliding window and the automatic location annotation of friction step fault provided by the embodiment of the present invention; the figure shows the abnormal area at the left end (25~65mm), the healthy benchmark area in the middle section (150~300 mm), and the severe abnormal area at the right end (375~425 mm, AUC peak value 71.8 A·mm). Detailed Implementation

[0052] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. This embodiment uses the X-axis and Y-axis linear feed system of a certain model of high-speed CNC drilling and tapping center (FANUC system, commonly known as the "Little Yellow Machine") as the diagnostic object. This machine tool has been running continuously for approximately 800 hours. Operators reported that when machining large workpieces, the X-axis occasionally produces abnormal noise at its right-hand limit position, and the machined surface shows vibration marks. However, conventional vibration testing did not reveal any obvious abnormalities. The method of this invention is used to accurately locate abnormal frictional force distribution along the entire stroke of the X-axis.

[0053] Physical excitation mechanism and NC instruction design logic:

[0054] To accurately elicit the latent frictional degradation characteristics of machine tools and provide a clean data flow for subsequent "instruction domain space mapping," this invention designs a set of stepped, segmented steady-state test G-code. The core design principles of this code are as follows:

[0055] Artificially implanted G04 isolation zone: During the switching intervals between different test conditions (e.g., before switching from high-speed positioning to low-speed crawling), the program forcibly inserts a G04 pause command with a time constant greater than 1000 ms. This command puts the servo system into a zero-speed lock state, forcibly cutting off residual inertial vibration and dynamic current overshoot, ensuring that the initial state of subsequent conditions is zero speed and zero dynamic load, providing a natural zero kinetic energy boundary for automatic state machine slicing.

[0056] Stepped low-speed steady-state crawling: To obtain pure frictional characteristics (rather than inertial forces), the G01 command is used to control the machine tool table to perform a full-stroke "reciprocating crawl" at an extremely low speed (10 mm / min, i.e., F10). F10 is chosen because at this speed, acceleration and deceleration inertial torques are negligible, and the servo motor's output current is almost entirely equal to the steady-state current required to overcome the guide rail friction and the leadscrew preload. This pure steady-state control flow, which eliminates acceleration and deceleration inertial torques, is the direct physical prerequisite for reconstructing the "space energy consumption hysteresis loop."

[0057] Segmented coverage of the entire stroke: To avoid cumulative errors that may be introduced by a single long-stroke reciprocating motion, the X-axis adopts a five-segment stepped design: 5 → 90 mm → quick positioning back to 5 mm → 5 → 180 mm → quick positioning back to 5 mm... and so on, up to 437 mm. After each segment, the cursor quickly returns to the starting point, ensuring that the starting point of each reciprocating motion is the same, facilitating the comparison of friction characteristics in different sections.

[0058] Hardware deployment and data acquisition (step S1):

[0059] Hardware configuration: Triaxial accelerometers are mounted on the X-axis and Y-axis worktable surfaces respectively. Magnetic mounts are used for adsorption, with a layer of standard A4 paper (approximately 0.1 mm thick) placed between the sensor mount and the machine tool's metal surface as an electrical insulation medium. This A4 paper serves to cut off common-mode interference transmitted from the machine tool's high-frequency chopper to the sensor housing via the machine tool body. All data acquisition devices are powered by independent lithium batteries and have no wire connection to the machine tool's electrical system, completely eliminating random electromagnetic crosstalk introduced by the power grid.

[0060] Data Acquisition: The actual current IEFF (unit: A, absolute value), speed command VCMD (unit: mm / min), and actual feedback position POSF (unit: mm) inside the servo driver are read at a frequency of 1000 Hz via the RS485 communication interface. The vibration sensor sampling rate is set to 10240 Hz. Simultaneously, the ambient temperature (25±1℃ in this embodiment) is recorded at a frequency of 1 Hz to eliminate the influence of temperature on the current baseline.

[0061] Operating conditions: Run the above-designed diagnostic G-code to collect three repeated experimental data (Rep1, Rep2, Rep3) for the X-axis (travel 5~437 mm, five-segment reciprocating motion) and Y-axis (travel 0~180 mm, single reciprocating motion). Each experiment was spaced 10 minutes apart to allow the guide rail temperature to return to room temperature.

[0062] Synchronous Alignment: Utilizing the G04+G01 impact markers (G04 X0.1 → G01 F10000X2) embedded at the beginning and end of the program, the maximum value of the first derivative of the torque command TCMD is extracted as the electrical anchor point T_elec. The point where the rising edge of the envelope of the triaxial composite vibration energy (the square root of Ax²+Ay²+Az²) exceeds 5 times the mean of the background noise is extracted as the mechanical anchor point T_mech. The delay Δt corresponding to the maximum value of the cross-correlation function is calculated; in this embodiment, Δt ≈ 1.22 ms. The vibration sequence is shifted forward by 1.22 ms, and zero-padding is applied to the boundaries. Then, cubic spline interpolation is performed on IEFF, VCMD, and POSF, upsampling to 10240 Hz to ensure complete alignment with the time axis of the shifted vibration signal. Finally, median filtering (window length 5) is applied to the interpolated signal to remove isolated spikes.

[0063] Working condition segmentation: A finite state machine was constructed based on VCMD to extract all LowSpeed ​​data segments that satisfy 5 mm / min < |VCMD| < 50 mm / min and a duration ≥ 0.1 seconds. For the X-axis, a total of 5 LowSpeed ​​semantic blocks were identified (each segment includes one forward and one reverse movement), with spatial intervals of 5~90 ​​mm, 90~180 mm, 180~270 mm, 270~360 mm, and 360~437 mm, respectively.

[0064] Spatial grid resampling and hysteresis loop comparative analysis (steps S2 and S3):

[0065] Spatial grid construction: The time independent variable t is completely removed. POSF and IEFF data are extracted from each LowSpeed ​​semantic block for forward motion (VCMD > 0) and reverse motion (VCMD < 0). A standard spatial grid S_grid is constructed, ranging from the minimum POSF to the maximum POSF of the segment with a spacing of 1 mm. Using linear interpolation, the non-uniformly sampled forward current values ​​in the time domain are mapped to each integer millimeter point on the S_grid to obtain the IEFF. - (s); similarly, we obtain IEFF. + (s).

[0066] Hysteresis loop drawing: IEFF - (s) and IEFF + (s) Plotted in the same coordinate system, with the horizontal axis representing s (mm) and the vertical axis representing IEFF (A). Forward current is represented by a solid blue line, and reverse current is represented by a dashed red line.

[0067] Analysis of health baseline (Y-axis) results:

[0068] As shown in Figure 2, during the entire stroke of the healthy Y-axis from 0 to 180 mm, the positive real current (blue solid line) and the reverse real current (red dashed line) form a very regular, flat, and almost parallel closed band. Specific values: the average positive current is approximately +0.52 A, and the average reverse current is approximately -0.81 A (the negative sign indicates direction).

[0069] The vertical spacing of the rings remains between 0.65 and 0.72 A, with a fluctuation range of less than 0.07 A. This indicates that the Y-axis guide rail assembly has extremely high parallelism, the slider is subjected to uniform force throughout its entire stroke, and the lubricating oil film coverage is continuous, indicating good hydrodynamic lubrication. Therefore, the Y-axis can be used as the health benchmark of this invention, and its AUC(s) curve (not given separately) is approximately horizontal throughout its entire stroke, with a mean of approximately 42 A·mm and a standard deviation of approximately 1.2 A·mm.

[0070] Analysis of hysteresis loop results across the entire stroke of the degraded axis (X-axis):

[0071] As shown in Figure 3, the hysteresis loop along the entire X-axis exhibits significant spatial non-uniformity:

[0072] In the 0~250 mm range: the forward current is stable at 0.75~0.85 A, the reverse current is stable at 0.95~-1.05 A, and the ring width is about 0.8~1.0 A, which is close to the healthy level of the Y-axis.

[0073] In the 250~300 mm range: the forward current slowly rises to 0.95 A, the reverse current drops to -1.15 A, and the ring width expands to about 1.2 A.

[0074] In the 300~437 mm range (right end): the forward current rises sharply from 0.95 A to 1.62 A (end), while the reverse current drops sharply from -1.15 A to -1.85 A (end). The ring width expands to over 2.0 A, approximately 100% larger than in the middle section. This "space expansion" directly indicates that the servo motor needs to output twice the torque at the right end to maintain a constant speed, proving the existence of significant additional resistance in this region.

[0075] Repeatability verification: The full-length hysteresis loops of X_Rep2 and X_Rep3 (due to file length limitations, Rep2 does not include the complete right end) show a consistent trend within the overlapping interval.

[0076] Analysis of local sub-segment hysteresis loops and stick-slip characteristics:

[0077] Based on the five-segment physical design of the X-axis, the full-stroke hysteresis loop is divided into five local sub-segments, corresponding to sub-figures 1 to 5 in Figure 4.

[0078] Segment 1 (5~90 mm): The current curve is dense and smooth, with a ring width of about 0.9 A. Both the forward and reverse curves are smooth parabolic shapes with no obvious abnormalities.

[0079] Segment 2 (90~180 mm): Same as above, ring width approximately 0.95 A, smooth.

[0080] Segment 3 (180~270 mm): The ring width is about 1.0 A, and it is still smooth.

[0081] Segment 4 (270~360 mm): The ring width increases significantly to about 1.5 A, and the blue solid line shows high-frequency sawtooth fluctuations (frequency about 12 Hz, peak-to-peak value about 0.05 A). Irregular burrs also appear on the red dashed line.

[0082] Segment 5 (360~437 mm): The ring width increases sharply to about 2.0 A, the peak-to-peak value of the high-frequency sawtooth fluctuation increases to 0.15~0.2 A, and the red dashed line shows obvious "dip" and "jump".

[0083] Tribological Interpretation: The high-frequency current spikes appearing in segments 4 and 5 are typical characteristics of the "stick-slip phenomenon" at low speeds. The physical mechanism is as follows: after the lubricating oil film breaks down, dry friction or boundary friction forms between the worktable and the guide rail. The static friction force is greater than the dynamic friction force, causing the worktable to initially "stick" under the motor's push, and the motor torque continues to rise. When the thrust exceeds the static friction force, the worktable suddenly "slips," causing positional error overshoot. The servo system quickly adjusts the current, forming a current spike; then it sticks again, and the cycle repeats. This mechanism manifests as high-frequency, irregular sawtooth-like jumps in the current waveform. The appearance of this characteristic in segments 4 and 5 in this embodiment proves that lubrication has failed in this section, and the friction state has degenerated into dry friction.

[0084] Sliding window AUC absolute value integration and automatic positioning (steps S4, S5):

[0085] Parameter settings: Spatial sliding window length W = 50 mm, step size Δd = 10 mm. The half-width of the window is 25 mm, meaning that each AUC value represents the average energy consumption within a 25 mm radius to the left and right of that point. The integration method uses the trapezoidal rule, with a spatial interpolation resolution of 0.1 mm (the current has been interpolated to a 0.1 mm grid before integration to improve accuracy).

[0086] Calculation results: As shown in Figure 5, plot the AUC(s) curve (horizontal axis: center position of the window in mm; vertical axis: AUC value, unit A·mm).

[0087] Left end region (25–65 mm): AUC value reaches 62.0 A·mm at 25 mm, then decreases to around 55 A·mm. This is about 30% higher than the mid-section baseline.

[0088] Mid-range region (150~300 mm): AUC values ​​are stable between 45 and 50 A·mm. Calculations were performed within the 150~300 mm range, with a mean μ = 47.2 A·mm and a standard deviation σ = 2.1 A·mm. A dynamic envelope threshold was set as μ + 3σ = 47.2 + 6.3 = 53.5 A·mm (red dashed line in Figure 5).

[0089] In the right-hand region (375–425 mm): the AUC value rises continuously from 59 A·mm, reaching 71.8 A·mm at 425 mm, exceeding the threshold by approximately 34%. This forms a steep, anomalous step region. Simultaneously, in the 405–425 mm range, the AUC value exceeds 70 A·mm, exceeding the mid-range baseline by approximately 52%.

[0090] Anomaly Detection: The system detected multiple consecutive windows with AUC values ​​> 53.5 A·mm in the 375~425 mm range on the right (actually 6 consecutive windows), and the corresponding local sub-segment hysteresis loop (segment 5) showed obvious expansion and high-frequency stick-slip burrs. Therefore, the system automatically output the diagnostic conclusion: "Severe frictional degradation exists in the X-axis physical coordinate range of 375 mm ~ 425 mm, and the lubricating oil film is ruptured. It is recommended to immediately check the guide rail surface and oil passage in this range." At the same time, the AUC value in the 25~65 mm range on the left exceeds the threshold, but no significant expansion or burrs are observed in the hysteresis loop, which is judged as a moderate anomaly (possibly due to slight carbon buildup), and a level 2 warning is output.

[0091] Verification comparison and on-site results:

[0092] Comparison with vibration signals: Under the same low-speed crawling conditions, the synchronously acquired triaxial composite vibration energy (RMS value) is shown in the lower subplot of Figure 4. The vibration amplitude throughout the entire stroke is less than 0.05 g, and there are no position-related fluctuations or peaks. If only vibration analysis is relied upon, a false diagnosis of "no abnormality on the X-axis" would be obtained, leading to missed diagnoses. However, this invention successfully located the millimeter-level friction degradation area. Field verification: Maintenance personnel, based on the diagnostic coordinates (375~425mm) output by this invention, removed the X-axis end protective cover and visually inspected it, finding that:

[0093] The surface of the guide rail has obvious black carbon deposits and drying marks in the range of about 380~420 mm, and feels rough to the touch;

[0094] The corresponding lubricating nozzle is blocked, and no oil can be squeezed out when using the grease gun, while the middle section nozzle can dispense oil normally;

[0095] There were minor scratches on the inside of the protective cover, but these were not the main source of resistance. Repair measures: Clean carbon deposits (using fine sandpaper to polish and then clean), unclog oil passages, refill with grease, and adjust the protective cover clearance. After completion, the diagnostic program was run again, and the hysteresis loop width on the right end of the X-axis returned to normal (forward current dropped to 0.92 A, reverse current -1.08 A), and the AUC peak value decreased to 52.0 A·mm (below the threshold of 53.5 A·mm).

[0096] This embodiment demonstrates that the present invention can:

[0097] ① In the low-speed range where vibration signals are completely ineffective, current signals are used for diagnosis;

[0098] ② The friction anomaly was located within ±1 mm accuracy (the actual anomaly range is 375~425 mm, the field verification range is about 380~420 mm, the error is only ±5 mm, which is far better than the tens of millimeters error of traditional methods).

[0099] ③ Quantitatively differentiate the severity of abnormalities (the proportion of AUC values ​​exceeding the threshold).

[0100] ④ Identify the micro-friction state (slippery burrs indicate oil film rupture).

Claims

1. A method for locating frictional degradation in a machine tool feed system based on spatial energy consumption hysteresis loop and sliding window AUC integral, characterized in that, Includes the following steps: S1. Collect external high-frequency vibration signals and internal servo control parameters under multiple working conditions of the machine tool. The internal servo control parameters include at least the real current signal IEFF, speed command VCMD, and feedback position POSF. Perform cross-correlation sub-millimeter level alignment using the physical impact abrupt changes of the internal and external signals. Construct a finite state machine based on the speed command VCMD to extract the low-speed crawling steady-state data segment where the command speed meets the preset low-speed crawling threshold range and the duration exceeds the stable threshold. S2. Remove the time variable t and extract the forward and reverse motion real current sequences under the low-speed crawling condition; construct a standardized spatial grid coordinate system with uniform accuracy, and use an interpolation algorithm to map the forward and reverse real currents onto the standardized spatial grid to obtain the forward spatial current function. and reverse space current function (s), where s are discrete physical space coordinates; S3, under the same spatial coordinates s (s) and (s) Perform geometric closure to construct a full-stroke bidirectional spatial energy consumption hysteresis loop; and cut and decouple the full-stroke hysteresis loop into several independent and continuous local sub-segment hysteresis loops according to preset rules; S4. Set a sliding window of length W and step size Δd; within each sliding window, calculate the absolute area AUC(s) enclosed by the forward and reverse current curves: S5. Compare the spatial energy area distribution curve, extract the mean and standard deviation of the AUC of the stable benchmark segment as the health benchmark, and set the dynamic envelope threshold. When the AUC(s) of a specific physical space interval continuously exceeds the dynamic envelope threshold, and the corresponding local sub-segment hysteresis loop undergoes significant longitudinal expansion or high-frequency jump, output the diagnostic physical coordinates of frictional degradation in the spatial interval.

2. The method according to claim 1, characterized in that, The preset low-speed crawling threshold range mentioned in step S1 is 5 mm / min < |VCMD| < 50 mm / min, and the duration exceeding the stability threshold is ≥0.1 seconds.

3. The method according to claim 1, characterized in that, The method for obtaining the physical impact abrupt change edge in step S1 is as follows: a combination instruction of G04 pause and G01 shift is implanted in the excitation program to generate an artificial physical impact waveform.

4. The method according to claim 1, characterized in that, The cross-correlation sub-millimeter alignment in step S1 specifically involves: extracting the maximum value of the first derivative of the servo torque command as the electrical anchor point, extracting the point where the rising edge of the synthetic vibration energy envelope exceeds the background noise threshold as the mechanical anchor point, calculating the delay offset through the cross-correlation function and performing vibration sequence shift compensation, and then upsampling and interpolating the intrinsic servo parameters.

5. The method according to claim 1, characterized in that, The accuracy of the standard spatial grid in step S2 is 0.1 mm to 1 mm, and the interpolation algorithm is linear interpolation or cubic spline interpolation.

6. The method according to claim 1, characterized in that, The preset rule mentioned in step S3 is: automatic cutting based on the machine tool physical feed clearance or preset spatial segment length.

7. The method according to claim 1, characterized in that, In step S4, the window length W is 50 mm, the step size Δd is 10 mm, and the integration is the trapezoidal integration method with a spatial interpolation resolution better than 0.5 mm.

8. The method according to claim 1, characterized in that, The dynamic envelope threshold mentioned in step S5 is μ + k・σ, where μ is the mean of AUC, σ is the standard deviation, and k is a preset coefficient, preferably k=3; the continuous breakthrough refers to at least 3 consecutive sliding windows exceeding the threshold; the significant expansion of the longitudinal span refers to the ring width increasing by more than 50% compared to the healthy benchmark; the high-frequency jump refers to the appearance of irregular sawtooth patterns with amplitude > 0.05A and frequency > 10Hz in the current waveform.

9. The method according to claim 1, characterized in that, When acquiring external high-frequency vibration signals in step S1, an electrical insulating medium is sandwiched between the sensor adsorption base and the metal surface of the machine tool, and both the external and internal acquisition devices are powered by independent DC power modules.

10. The method according to claim 9, characterized in that, The electrical insulation medium is standard A4 paper or polyimide film with a thickness of 0.05 mm to 0.15 mm.

Images