A method and system for dynamic compensation of wear error of a tilt centerless grinding machine

Abstract

This application relates to the field of precision grinding control, and particularly to a method and system for dynamic compensation of wear error in a tilting centerless grinder. The method includes: acquiring infrared temperature images of the grinding arc zone; calculating the spatial temperature gradient modulus and the positive temperature rise value to construct a thermal shock intensity fusion factor; generating an adaptive weighting coefficient based on the local thermal information entropy deviation, and integrating it with the fusion factor in a spatiotemporal manner to obtain an equivalent thermal shock dose factor; calculating the increment of the equivalent thermal shock dose factor for adjacent workpieces, and generating a softening degree asymptotic coefficient through historical truncated standard deviation outlier clipping and exponential damping accumulation; and calculating and generating a dynamic wear offset based on the softening degree asymptotic coefficient and conventional wear reference quantity through saturation limiting mapping, and superimposing it to form a dynamic compensation command. This application has the technical effect of accurately quantifying the thermal softening trend, suppressing interference, and ensuring that the compensation command is safe and within tolerance, significantly improving the accuracy of long-batch machining.

Description

Technical Field

[0001] This application relates to the field of precision grinding control, and in particular to a method and system for dynamic compensation of wear error in a tilting centerless grinder. Background Technology

[0002] Inclined centerless grinders are widely used in continuous through-grind grinding of brittle and hard workpieces such as cemented carbide. During grinding, a high-density heat flow is generated in the contact arc area between the grinding wheel and the workpiece in a very short time, causing transient local temperature rise and thermal softening of the grinding wheel bond, which in turn causes changes in the radial dimension of the grinding wheel and grinding errors. However, the random obstruction of the atomized layer of water-based grinding fluid and the abrupt change in the temperature field at the inlet and outlet of the contact arc introduce regional noise and truncation artifacts into infrared thermal imaging monitoring, making it difficult to accurately extract the true thermal shock intensity. At the same time, the mechanical wear and thermal softening of the grinding wheel in long batch processing are coupled. Traditional compensation methods mostly rely on fixed empirical benchmarks or offline calibration, which cannot track thermal load fluctuations and bond state drift in real time. This can easily cause oscillations or overcompensation in compensation commands, ultimately leading to workpiece dimensional deviations and decreased surface consistency, making it difficult to meet the real-time dynamic compensation requirements of modern high-precision and automated grinding processes. Summary of the Invention

[0003] To address the technical problems of inaccurate quantification of grinding thermal shock and easy oscillation of compensation commands, this application provides a dynamic compensation method and system for wear error of a tilting centerless grinder.

[0004] In a first aspect, this application provides a method for dynamic compensation of wear error in a tilting centerless grinder, employing the following technical solution, including: Infrared temperature images of the grinding arc region are acquired, and the spatial temperature gradient modulus and positive temperature rise value of each pixel are calculated. A thermal shock intensity fusion factor is constructed based on the spatial temperature gradient modulus and positive temperature rise value. For the grinding wheel contact arc region at each moment, the local thermal information entropy deviation of the grinding wheel contact arc region is calculated based on the infrared temperature image, and an adaptive weighting coefficient is generated using the local thermal information entropy deviation. The thermal shock intensity fusion factor and the adaptive weighting coefficient are integrated over the contact arc area and the processing duration to obtain the equivalent thermal shock dose factor of the current workpiece. The increment of the equivalent thermal shock dose factor of the current workpiece and its adjacent previous workpiece is calculated. The increment of the equivalent thermal shock dose factor is clipped based on the truncated standard deviation of the historical equivalent thermal shock dose factor increment, and exponential damping accumulation is applied to the clipped dose factor increment to generate a softening degree asymptotic coefficient. Based on the softening degree asymptotic coefficient and a preset conventional wear reference amount, a dynamic wear offset is generated by saturation limiting mapping, and it is superimposed on the conventional wear reference amount to form a dynamic compensation command for subsequent workpiece grinding.

[0005] Optionally, an infrared thermal imager can be used to continuously acquire two-dimensional temperature images in the infrared band of the grinding arc area at a preset frame rate.

[0006] Optionally, the spatial temperature gradient modulus and the positive heating value within the contact arc region of the current frame image are used to normalize the spatial temperature gradient modulus and the positive heating value respectively, and the normalized results are fused to obtain the thermal shock intensity fusion factor.

[0007] Optionally, a uniform heating reference entropy value is calculated based on the local thermal information entropy of a uniformly heated candidate region whose spatial temperature gradient modulus is lower than the median in the current frame temperature image; the adaptive weighting coefficient is generated using the exponential function of the local thermal information entropy deviation.

[0008] Optionally, the grinding zone entry time is used as the lower limit of integration and the grinding zone exit time is used as the upper limit of integration. The area integral is performed on the grinding wheel contact arc region, and the grinding duration is performed as the time integral. The integral term is the product of the adaptive weighting coefficient and the thermal shock intensity fusion factor.

[0009] Optionally, the truncation standard deviation of historical increments is calculated using the effective dose factor increments of the most recent several effective workpiece cycles after removing identified outliers; an increment pruning threshold coefficient is set, and if the absolute value of the current increment is greater than the product of the threshold coefficient and the truncation standard deviation, it is determined to be an outlier, and the outlier is replaced by the most recent effective increment value.

[0010] Optionally, the current trimmed dose factor increment after smoothing is normalized to the current conventional wear reference amount, and then the normalized increment is superimposed on the softening degree asymptotic coefficient of the previous workpiece to obtain the softening degree asymptotic coefficient of the current workpiece; wherein, the initial softening degree asymptotic coefficient is zero, and as the conventional wear reference amount increases, the growth rate of the softening degree asymptotic coefficient caused by the same increment decreases.

[0011] Optionally, the saturation limiting mapping calculation uses a hyperbolic tangent function, generates input variables based on the softening degree asymptotic coefficient and the conventional wear reference amount, and multiplies the output value of the hyperbolic tangent function by a preset saturation upper limit to obtain the dynamic wear offset; the saturation upper limit is determined according to the dimensional tolerance of the workpiece; when the softening degree asymptotic coefficient increases, the dynamic wear offset approaches and does not exceed the saturation upper limit at a decreasing rate.

[0012] Secondly, this application provides a dynamic wear error compensation system for an inclined centerless grinder, which adopts the following technical solution: A dynamic wear error compensation system for an inclined centerless grinder includes a processor and a memory. The memory stores computer program instructions, which, when executed by the processor, implement the aforementioned dynamic wear error compensation method for an inclined centerless grinder.

[0013] This application has the following technical effects: 1. By fusing the spatial temperature gradient modulus and the positive heating value to construct a thermal shock intensity fusion factor, and combining it with the local thermal information entropy deviation to generate an adaptive weighting coefficient, the noise interference caused by grinding fluid atomization scattering and contact arc boundary artifacts is effectively suppressed. This achieves accurate extraction and weighted integration of the thermal shock intensity of the real heated area, providing a pure and reliable physical quantity basis for subsequent thermal softening assessment.

[0014] 2. An outlier clipping mechanism based on the historical incremental truncation standard deviation and an exponential damping accumulation algorithm are introduced to generate asymptotic coefficients of softening degree. This not only eliminates atypical thermal shock jumps caused by occasional events such as instantaneous interruption of grinding fluid and abnormal accumulation of grinding debris, but also incorporates the physical modulation law of wear amount sensitivity to thermal softening. This avoids step distortion and high-frequency oscillation of compensation commands, making the softening trend estimation highly smooth and consistent with the actual physical aging process of grinding wheels.

[0015] 3. The dynamic wear offset is calculated using a hyperbolic tangent mapping with saturation limiting characteristics, which strictly limits the compensation increment within the safe process boundary naturally derived from the workpiece dimensional tolerance. This fundamentally eliminates the problem of overcompensation and out-of-tolerance due to thermal softening accumulation in the later stages of long batch processing. At the same time, the online adaptive superposition of conventional wear reference and dynamic offset realizes a data-driven feedforward compensation closed loop that eliminates the need for manual calibration, which significantly improves the long-term machining accuracy and operating efficiency of the grinding machine. Attached Figure Description

[0016] Figure 1 This is a flowchart of a method for dynamic compensation of wear error in a tilting centerless grinder. Detailed Implementation

[0017] This application provides a method and system for dynamic compensation of wear error in a tilting centerless grinder.

[0018] This application discloses a method for dynamic compensation of wear error in a tilting centerless grinder, referring to... Figure 1 ,include: S1: Obtain the infrared temperature image of the grinding arc area, calculate the spatial temperature gradient modulus and positive temperature rise value of each pixel, and construct the thermal shock intensity fusion factor based on the spatial temperature gradient modulus and positive temperature rise value.

[0019] When brittle and hard workpieces such as cemented carbide undergo continuous through-grind grinding on a tilting centerless grinder, a high-density heat flow is generated in the contact arc area between the grinding wheel and the workpiece in a very short time, leading to a transient local temperature rise in the bond. However, the random obstruction of the atomized layer of the water-based grinding fluid and the abrupt temperature field changes at the boundary of the contact arc introduce regional noise and truncation artifacts into the temperature image output by the infrared thermal imager. Since the subsequent construction of the thermal shock intensity factor directly depends on the accurate extraction of the spatial gradient and time derivative of the temperature field, the undifferentiated full-image gradient calculation will include the random fluctuations of pixel values ​​caused by atomization scattering and the high-frequency components at the steep boundary transition in the thermal shock evidence, thus generating a large number of pseudo-intensity point sets that are unrelated to the actual material removal heat effect. Therefore, this step first continuously acquires two-dimensional temperature images in the infrared band of the grinding arc area using an infrared thermal imager. Let the two-dimensional temperature image acquired at time t be... , It is the gray value of the pixel at coordinates (x, y) in the image at time t. The coordinates are parallel to the axis of the grinding wheel. The coordinates are along the tangent direction of the grinding arc. For time, the wavelength range of the mid-infrared band is 3μm to 5μm. The acquisition frame rate of the infrared thermal imager is determined by its hardware capabilities and only needs to meet the time resolution requirements for the transient characteristics of grinding thermal shock; it does not constitute an adjustable parameter in the method.

[0020] For each frame of the temperature image, the spatial temperature gradient is calculated pixel-by-pixel to extract the spatial temperature gradient modulus of the thermal shock front on the contact arc plane. The spatial temperature gradient modulus is defined. This is the square root of the sum of the squares of the partial derivatives of the temperature along the axial and tangential directions of the grinding wheel, i.e. in, For pixels At any moment The spatial temperature gradient modulus, expressed in degrees Celsius per millimeter (°C / mm); For temperature Along the axis of the grinding wheel The partial derivatives; For temperature Coordinates along the tangent direction of the grinding arc The partial derivatives of .

[0021] In this embodiment, the spatial temperature gradient modulus directly characterizes the local rate of temperature change at that pixel along the two-dimensional grinding arc surface: The larger the value, the steeper the spatial temperature change at that point, corresponding to a stronger thermal shock front spatial interface; conversely, the smaller the spatial temperature gradient modulus, the gentler the temperature change and the weaker the spatial characteristics of thermal shock.

[0022] Meanwhile, since the nature of grinding thermal shock is a sudden rise in temperature rather than a slow drop, in order to eliminate the non-physical contribution of the cooling tail stage to the calculation of thermal shock intensity, only the positive component of temperature change over time is retained, and the positive heating value is defined. The non-negativity of the temperature partial derivative with respect to time: in, For pixels At any moment The positive temperature rise value, expressed in degrees Celsius per millisecond (°C / ms); For temperature Regarding time The partial derivatives; This indicates that only non-negative values ​​are taken.

[0023] The larger the value, the more intense the heating process is at that point, and the higher the potential for thermodynamic damage. If the value is zero, it indicates that the point is in a cooling or isothermal phase and does not contribute to the thermal shock intensity. This decouples the transient heating rate and cooling recovery process at each point.

[0024] To eliminate the dimensional difference between spatial gradient modulus and heating rate, and to ensure the comparability of fusion intensity under varying conditions, a normalization benchmark is determined in real time using the statistical characteristics of each frame. For the first... Frame image, calculate the values ​​of all pixels within the contact arc region. The median, denoted as the reference spatial gradient modulus of this frame. (If the median is zero, take the smallest non-zero value); similarly, calculate the median for all pixels. The median, denoted as the reference heating rate for this frame. Divide the two fundamental quantities by their respective intra-frame references to convert them into dimensionless relative intensities, then fuse them in an equally weighted manner to construct a thermal shock intensity fusion factor. in, For pixels At any moment The thermal shock intensity fusion factor is dimensionless. This is the spatial temperature gradient modulus at that point; For the first The reference spatial gradient modulus of a frame is determined by all pixels within the contact arc of that frame. The median is determined; This represents the positive temperature rise value at that point; For the first The reference heating rate of a frame is determined by all pixels within the contact arc of that frame. The median is determined.

[0025] The larger the value, the more pronounced the spatial temperature change (high gradient region) and transient temperature rise characteristics of the pixel relative to the entire image frame, and therefore the stronger the actual thermal shock experienced by that point; when When the value is close to zero, it indicates that there is neither a significant spatial front nor a rapid temperature rise at that point, and the thermal shock is extremely weak or it is in an artifact zone. Because the reference value is determined independently for each frame, when the grinding wheel passivation causes changes in the overall temperature level or cooling conditions, the reference automatically follows, and the fusion factor always reflects the relative thermal shock intensity under the current operating conditions, completely replacing the contribution coefficient that originally needed to be calibrated offline.

[0026] Thus, a dimensionless thermal shock intensity fusion factor has been obtained for each pixel in each frame of the temperature image. This factor, as a characterization of the intensity of local thermal shock, retains the spatial and temporal differential features closely related to transient thermal shock in the original temperature field, providing a reliable and adaptive intermediate input for subsequent integration to generate the equivalent thermal shock dose factor.

[0027] S2: For the grinding wheel contact arc area at each moment, calculate the local thermal information entropy deviation of the grinding wheel contact arc area based on the infrared temperature image, and use the local thermal information entropy deviation to generate an adaptive weighting coefficient. Integrate the thermal shock intensity fusion factor and the adaptive weighting coefficient over the contact arc area and processing duration to obtain the equivalent thermal shock dose factor of the current workpiece.

[0028] Because the scattering and absorption of the atomized layer of water-based grinding fluid in the infrared band will produce regional random fluctuations in the temperature image, and the abrupt temperature field change at the workpiece-grinding wheel contact arc entry and exit points will introduce boundary truncation artifacts in the image, these non-real grinding thermal effect regions affect the thermal shock intensity fusion factor. This manifests as a set of pseudo-extreme points on the same or even higher order of magnitude as the actual thermal shock region. If all pixels... Equal-weighted integration, due to spurious intensities caused by fogging and boundary disturbances, degrades the signal-to-noise ratio of the final extracted equivalent thermal shock dose factor. Therefore, this step, for each time step... grinding wheel contact arc area The system distinguishes between the interference area and the real heated area based on the degree of disorder in the local neighborhood of the temperature image, and generates adaptive weighting coefficients accordingly.

[0029] Regarding the first Each pixel located within the contact arc region in the frame temperature image. Take a local neighborhood window centered on that point. For example, the local neighborhood window can be... A pixel matrix.

[0030] Perform histogram analysis on the temperature values ​​of all pixels within the window and calculate the local thermal entropy. Its definition is in, For pixels At any moment The local thermal entropy is dimensionless. The temperature value within the window falls into the first Probability density estimation for each quantization interval This represents the total number of quantization intervals for the histogram statistics.

[0031] Local thermal entropy quantifies the disorder of temperature distribution within a neighborhood. The larger the value, the more chaotic and disordered the temperature distribution in the neighborhood, and the more obvious the characteristics of fogging scattering or abrupt boundary changes. The smaller the value, the more uniform and orderly the temperature distribution, and the more likely it is to be a true heating region with stable thermal shock.

[0032] To obtain an entropy benchmark for a uniformly heated state online, no prior calibration is required. In the... In the frame image, calculate the values ​​of all pixels within the contact arc. The median of the gradient modulus is used to select pixels with gradient moduli lower than the median as candidate regions for uniform heating. The arithmetic mean of the local thermal entropy of these candidate region pixels is used as the uniform heating reference entropy value for this frame. Then, the local thermal entropy deviation of each pixel is calculated: in, For pixels At any moment The local thermal information entropy deviation is dimensionless. This represents the local thermal entropy at that point. For the first The uniform heating reference entropy value of the frame.

[0033] The larger the value, the further the temperature distribution pattern of the neighborhood of the pixel deviates from the uniform heating state, and the higher the probability that the point belongs to the fog scattering or boundary artifact region. When the value approaches zero, the point is very likely located within the actual heated zone.

[0034] Based on the local thermal information entropy deviation, the adaptive weighting coefficient is directly given by the following formula: in, For pixels At any moment The adaptive weighting coefficients are dimensionless and range from 1 to 1. ; For the natural constant An exponential function with base 0; This represents the local thermal information entropy deviation at that point.

[0035] The closer the value is to 1, the closer the point is to a uniform heating state, and its thermal shock intensity fusion factor will be fully preserved in subsequent integrations; The closer the value is to 0, the more likely the point is to be a disturbance region, and its contribution to thermal shock intensity will be significantly suppressed. When When the weight increases from 0 to 2, it rapidly decreases from 1 to approximately 0.14, achieving strong suppression of artifact regions; when With a constant time weight of 1, information about the actual heated area passes through without loss. This natural exponential form provides sufficient discrimination within the typical dynamic range of information entropy, without the need for an additional sensitivity coefficient.

[0036] Capture every moment After applying a weighted matrix to the entire contact arc region, the adaptive weighting coefficients of all pixels are multiplied by the thermal shock intensity fusion factor at the same location, and spatial two-dimensional area integration and temporal integration are performed to calculate the equivalent thermal shock dose factor of the current workpiece. The mathematical form of this integral is: in, The equivalent thermal shock dose factor for the current workpiece is dimensionless. This refers to the moment when the front end of the workpiece enters the grinding zone. The moment when the tail end of the workpiece exits the grinding zone is the integral interval covering the entire grinding duration of the workpiece. The contact arc region of the grinding wheel is determined in real time by the grinding depth and the geometric parameters of the grinding wheel. These are adaptive weighting coefficients; This is the thermal shock intensity fusion factor.

[0037] This integral operation sums the weighted thermal shock intensity over the entire contact arc surface in space and accumulates it over the entire grinding duration of the workpiece in time, thereby aggregating the transient, distributed pixel-level thermal shock characterization into a single physical quantity. The larger the value, the stronger the total thermal shock the workpiece experiences throughout the grinding cycle, and the greater the driving force for the thermal softening of the grinding wheel bond; conversely, the smaller the value, the stronger the thermal shock. A smaller value indicates a lighter heat load. Because the adaptive weighting coefficient effectively suppresses fogging and boundary artifacts, the generated... It only reflects the intensity, spatial range and duration of thermal shock in the actual heated area, providing a pure measure of heat load for subsequent softening estimation.

[0038] This completes the equivalent thermal shock dose factor for the current workpiece. The calculation yielded a workpiece-level total thermal shock index that is robust to accidental disturbances, which can be directly used for softening trend analysis between adjacent workpieces.

[0039] S3: Calculate the equivalent thermal shock dose factor increment of the current workpiece and its adjacent previous workpiece, perform outlier clipping on the equivalent thermal shock dose factor increment based on the truncated standard deviation of the historical equivalent thermal shock dose factor increment, and apply exponential damping accumulation to the clipped dose factor increment to generate asymptotic softening coefficient.

[0040] After the equivalent thermal shock dose factor for a single workpiece is generated, this dose factor sequence does not exhibit a stable progression during long batch machining. Due to non-monotonic abrupt changes in the dose factor caused by binder fatigue at the end of the wheel dressing interval, and random walks caused by instantaneous fluctuations in grinding fluid flow and brief accumulation of grinding debris, the dose factor increments calculated for adjacent workpieces will contain outliers unrelated to the actual thermal softening trend. Directly accumulating these increments containing outliers will amplify random disturbances, leading to oscillations in the compensation command. Therefore, for the first... The workpiece has been processed and its equivalent thermal shock dose factor has been obtained. This state is first calculated by determining its increment compared to the previous workpiece dose factor: in, For the first The workpiece is relative to the first The increment of the equivalent thermal shock dose factor for each workpiece, dimensionless; For the first Equivalent thermal shock dose factor for each workpiece; This is the equivalent thermal shock dose factor for the previous workpiece. For the first workpiece in the batch, .

[0041] A positive value indicates that the current workpiece experiences more thermal shock than the previous one, and the tendency for the grinding wheel to soften may accelerate; a negative value indicates the opposite. This increment of the equivalent thermal shock dose factor directly reflects the change in thermal load between adjacent processing cycles.

[0042] Since atypical thermal shock jumps can be caused by occasional events such as instantaneous interruption of grinding fluid and abnormal accumulation of grinding debris, it is necessary to introduce an outlier pruning mechanism so that the equivalent thermal shock dose factor increment sequence retains only the normal fluctuations that reflect the slow drift of the true thermal state, thus avoiding step distortion in subsequent softening estimation.

[0043] To distinguish between normal fluctuations and outliers, a truncated standard deviation based on the historical equivalent thermal shock dose factor increment is introduced. The clipping mechanism involves constructing a sliding window of a preset length with a step size of 1 on the equivalent thermal shock dose factor increment. The statistical standard deviation of all equivalent thermal shock dose factor increments within this sliding window is then calculated. If the current equivalent thermal shock dose factor increment absolute value Greater than If the value is found to be out of bounds, it is replaced with the most recently occurring effective increment value to obtain the dose factor increment after trimming. Otherwise, retain the original value, i.e. The three-standard-deviation criterion is a standard statistical setting.

[0044] For example, the preset length of the sliding window can be set to 40 data points, i.e., 40 dose factor increments after clipping. The sliding window needs to be long enough to support reliable calculations of the truncated standard deviation and Otsu's method. A sample size ≥30 can avoid distortion of the Otsu threshold. At the same time, it needs to cover the typical slow-change cycle of wheel thermal softening to avoid fitting the trend with noise. However, it cannot be too long to avoid response lag or increased computational burden. For example, the preset length of the sliding window is set to 40 because this value satisfies both statistical estimation experience and covers the typical processing cycle of 7~20 minutes for a centerless grinder (corresponding to the thermal equilibrium transition period after wheel dressing). In actual deployment, the length can be set to 50~60 for high-noise conditions within 20~60 workpiece cycles to enhance anti-interference, 20~30 for fast-paced and high-sensitivity conditions to facilitate rapid tracking, and 35~45 for normal steady-state processing to balance smoothness and response.

[0045] It should be noted that the effective workpiece cycle is the workpiece cycle in which a complete grinding process is completed and the increment of its equivalent thermal shock dose factor is confirmed as a non-outlier value after outlier determination. If the sliding window length does not reach the preset number of values ​​at the beginning of the processing stage, outlier clipping is skipped and execution continues directly until the sliding window length reaches the preset number of values, at which point outlier clipping begins.

[0046] Directly accumulating the dose factor increments after clipping may still result in a coarse softening coefficient trajectory due to residual small perturbations. Therefore, adaptive smoothing is applied to the clipped dose factor increments. The smoothing coefficient is automatically adjusted based on the volatility of the increment sequence: the greater the volatility, the smaller the smoothing coefficient, to accelerate the tracking of the true trend; the smaller the volatility, the larger the smoothing coefficient, to obtain a highly smooth output. Specifically, within the sliding window, the Otsu method is used to traverse the clipped dose factor increments in chronological order within the sliding window, obtaining a first and second class with chronological order. The standard deviation of the second class is taken as the short-term standard deviation. The standard deviation of all post-clipping dose factor increments within the sliding window is taken as the long-term standard deviation. The smoothing coefficient is defined as follows: in, For smoothing coefficients; Short-term standard deviation; The standard deviation is the long-term standard deviation.

[0047] When the process is stable and the fluctuation of the dose factor increment after trimming is small... Approaching 1, the filter outputs a highly smooth sequence of trimmed dose factor increments with strong inertia, resulting in a smooth and gradual softening coefficient curve that conforms to the high inertia characteristics of the grinding wheel thermal softening physical process. However, when grinding fluid pulsation or slight wheel clogging occurs, short-term fluctuations in the trimmed dose factor increments intensify. The filter automatically decreases to near zero, reducing its inertia and responding more quickly to changes in the actual trend, thus avoiding the softening coefficient lagging behind the actual thermal state. This adaptive mechanism ensures that the smoothing intensity always matches the current processing stability.

[0048] The recursive formula for the filter increment is: in, For the first The filter increment corresponding to each workpiece is dimensionless. For smoothing coefficients; The filter increment for the previous workpiece, initial value ; This represents the dose factor increment after trimming.

[0049] It is the result of low-pass filtering of the equivalent thermal shock dose factor increment sequence, which removes high-frequency random residuals and retains the slowly varying components that reflect the thermal softening trend. The closer it is to 1, the smaller the contribution of the current increment to the filter output, and the more the output depends on historical accumulation; The closer it is to 0, the higher the proportion of the current increment being adopted immediately, and the faster the output response.

[0050] The sensitivity of grinding wheel bond thermal softening to mechanical wear decreases as the grinding wheel wears down. To ensure the softening coefficient reflects this physical law, the filter increment is normalized using the current standard wear value as the denominator before accumulation. (Softening degree asymptotic coefficient) Update as follows: in, For the first The asymptotic coefficient of softening degree after the workpiece is processed, dimensionless; This is the asymptotic coefficient of the softening degree of the previous workpiece, with an initial value. ; This is the filter increment; For the first The standard wear reference value for a workpiece before grinding is expressed in micrometers (μm). It serves as a unit length reference and is used only to make the denominator dimensionless; To prevent zero small constants.

[0051] Softening coefficient It monotonically increases with the accumulation of positive filter increments, reflecting the deepening softening of the grinding wheel binder under historical thermal shock. Under the same filter increment, when... When the values ​​are relatively small (low grinding wheel wear, abundant bonding agent), the denominator is relatively small. The significant increase corresponds to the fact that newly dressed grinding wheels are more sensitive to thermal shock; when When the value is large (grinding wheel is severely worn and little binder remains), the denominator increases, and the same filter increment results in... The significant decrease in the rate of increase reflects the physical effect of thermal softening tending to saturate in the later stages of wear.

[0052] The thermal softening of the grinding wheel bond is driven by the accumulation of the equivalent thermal shock dose factor. The conventional wear reference amount is introduced into the calculation of the softening degree progression coefficient to reflect the decay law of thermal softening sensitivity after the bond is continuously consumed by mechanical wear: the smaller the wear amount, the more sufficient the effective volume of the bond, and the greater the softening increase corresponding to the unit thermal shock increment; the larger the wear amount, the less the remaining volume of the bond that can participate in thermal softening, and the lower the marginal increase of softening under the same thermal shock, so that the gradual process of softening degree conforms to the actual physical state of the grinding wheel throughout its entire life cycle.

[0053] At this point, the gradual coefficient of softening degree has been completed. The recursive calculation yields a grinding wheel thermal softening metric that removes outliers, has adaptive smoothing, and incorporates wear sensitivity attenuation, which can be directly used to generate subsequent compensation offsets.

[0054] S4: Based on the softening degree progressive coefficient and the preset conventional wear reference amount, a dynamic wear offset is generated by saturation limiting mapping, and then superimposed on the conventional wear reference amount to form a dynamic compensation command for subsequent workpiece grinding.

[0055] To obtain the asymptotic coefficient of the softening degree corresponding to the current workpiece Subsequently, if an unbounded linear mapping is used, excessive compensation values ​​will be generated in the later stages of long batch processing, causing the workpiece dimensions to exceed tolerances. To address this, a hyperbolic tangent mapping function with a saturation upper limit characteristic is constructed. Its saturation upper limit is automatically determined by the workpiece dimensional tolerance, and the softening coefficient itself has embedded the modulation of softening sensitivity by wear amount.

[0056] Dynamic wear offset The calculation formula is: in, For the first The dynamic wear offset applied to each workpiece, in micrometers (μm); The upper limit of the offset saturation is taken as the upper specification limit of the workpiece diameter. With nominal size The absolute value of the difference, in μm; The function is the hyperbolic tangent, and its output range is... ; This is the asymptotic coefficient of the current workpiece's softening degree, dimensionless.

[0057] when When smaller, , Nearly linear growth, compensating for and sensitively following the small amount of additional wear in the early stages of thermal softening; as... Increase The function gradually saturates. The growth rate is decreasing and gradually approaches zero. This saturation characteristic ensures that, regardless of the extent of thermal softening accumulation, the compensation amount is always limited to the safe process boundary. Within this range, we must prevent deviations caused by overcompensation. Because... It is already embedded in step S3. Modulation of softening accumulation, the same thermal shock increment generated in the later stages of wear. Smaller growth, therefore The input can be automatically maintained within a reasonable dynamic range throughout the entire life cycle without the need for an additional mapping slope adjustment factor.

[0058] Conventional wear reference quantity For the first The radial mechanical wear of the grinding wheel before grinding each workpiece is automatically determined and updated during the machining process. Initial values ​​are obtained directly from the dimensional error of the first trial cut before batch processing. In subsequent processing, the workpiece diameter is periodically sampled, and linear regression is performed on the dimensional error sequence to obtain the average mechanical wear rate per unit workpiece. (μm / piece), and according to Recursive update. Therefore... The entire process requires no manual preset or offline calibration, and it adaptively tracks the actual wear process of the grinding wheel, providing an accurate benchmark for softening accumulation and compensation commands.

[0059] Finally, the dynamic wear offset is added to the conventional wear reference value to obtain the value for the first wear. Dynamic compensation instructions for each workpiece: in, For use in the first The dynamic compensation command for workpiece grinding is in micrometers (μm). For the first Standard wear reference value for each workpiece before machining; According to the first Dynamic wear offset generated by the thermal softening state of a workpiece.

[0060] This value combines basic compensation for grinding wheel mechanical wear with additional compensation caused by thermal softening. A higher value indicates that the feed axis needs to be offset more towards the workpiece to compensate for grinding wheel wear. Because... Constrained by the natural tolerances of the workpiece, It remains within the physical safety range. This command is converted into the position offset value of the grinding machine feed axis and preset in the servo control loop before subsequent workpiece grinding, forming feedforward compensation.

[0061] This completes the dynamic compensation command for the next workpiece. The generation of this data enables a closed-loop dynamic compensation system for wear errors, automatically driven by processing data.

[0062] This application also discloses a dynamic wear error compensation system for a tilting centerless grinder, including a processor and a memory. The memory stores computer program instructions, and when the computer program instructions are executed by the processor, a dynamic wear error compensation method for a tilting centerless grinder according to this application is implemented.

[0063] A dynamic wear error compensation system for an inclined centerless grinder also includes other components well known to those skilled in the art, such as a communication bus and a communication interface. Their settings and functions are known in the art and will not be described in detail here.

[0064] In this application, the aforementioned memory can be any tangible medium that contains or stores a program that can be used or combined with an instruction execution system, apparatus, or device. For example, a computer-readable storage medium can be any suitable magnetic or magneto-optical storage medium, such as resistive random access memory, dynamic random access memory, static random access memory, etc., or any other medium that can be used to store required information and can be accessed by an application program, module, or both.

[0065] The above are all preferred embodiments of this application, and are not intended to limit the scope of protection of this application. Therefore, all equivalent changes made in accordance with the structure, shape and principle of this application should be covered within the scope of protection of this application.

Claims

1. A method for dynamic compensation of wear error in a tilting centerless grinder, characterized in that, Including the following steps: Infrared temperature images of the grinding arc region are acquired, and the spatial temperature gradient modulus and positive temperature rise value of each pixel are calculated. A thermal shock intensity fusion factor is constructed based on the spatial temperature gradient modulus and positive temperature rise value. For the grinding wheel contact arc region at each moment, the local thermal information entropy deviation of the grinding wheel contact arc region is calculated based on the infrared temperature image, and an adaptive weighting coefficient is generated using the local thermal information entropy deviation. The thermal shock intensity fusion factor and the adaptive weighting coefficient are integrated over the contact arc area and processing duration to obtain the equivalent thermal shock dose factor of the current workpiece. Calculate the equivalent thermal shock dose factor increment of the current workpiece and its adjacent previous workpiece. Based on the truncated standard deviation of the historical equivalent thermal shock dose factor increment, perform outlier clipping on the equivalent thermal shock dose factor increment. Apply exponential damping accumulation to the clipped dose factor increment to generate asymptotic softening coefficient. Based on the softening degree progressive coefficient and the preset conventional wear reference amount, a dynamic wear offset is generated by saturation limiting mapping, and then superimposed on the conventional wear reference amount to form a dynamic compensation command for subsequent workpiece grinding.

2. The method for dynamic compensation of wear error in a tilting centerless grinder according to claim 1, characterized in that, The acquisition of the infrared temperature image of the grinding arc region includes: Two-dimensional temperature images in the infrared band of the grinding arc area are continuously acquired using an infrared thermal imager at a preset frame rate.

3. The method for dynamic compensation of wear error in a tilting centerless grinder according to claim 1, characterized in that, The step of constructing a thermal shock intensity fusion factor based on the spatial temperature gradient modulus and the positive heating value includes: By utilizing the statistical characteristics of the spatial temperature gradient modulus and the positive temperature rise value within the contact arc region of the current frame image, the spatial temperature gradient modulus and the positive temperature rise value are normalized respectively, and the normalized results are fused to obtain the thermal shock intensity fusion factor.

4. The method for dynamic compensation of wear error in a tilting centerless grinder according to claim 1, characterized in that, The calculation of the local thermal information entropy deviation of the grinding wheel contact arc region based on the infrared temperature image, and the generation of adaptive weighting coefficients using the local thermal information entropy deviation, includes: Based on the local thermal information entropy of the uniformly heated candidate region whose spatial temperature gradient modulus is lower than the median in the current frame temperature image, the uniformly heated reference entropy value is calculated; the adaptive weighting coefficient is generated using the exponential function of the local thermal information entropy deviation.

5. The method for dynamic compensation of wear error in a tilting centerless grinder according to claim 1, characterized in that, The step of integrating the thermal shock intensity fusion factor with the adaptive weighting coefficient over the contact arc area and processing duration includes: Using the entry time of the grinding zone as the lower limit of integration and the exit time of the grinding zone as the upper limit of integration, the area integral is performed on the contact arc region of the grinding wheel, and the time integral is performed on the grinding duration. The integral term is the product of the adaptive weighting coefficient and the thermal shock intensity fusion factor.

6. The method for dynamic compensation of wear error in a tilting centerless grinder according to claim 1, characterized in that, The outlier pruning of the equivalent thermal shock dose factor increment based on the truncated standard deviation of historical increments includes: Calculate the truncation standard deviation of historical increments using the effective dose factor increments from the most recent several effective workpiece cycles after removing identified outliers; set an increment pruning threshold coefficient, and if the absolute value of the current increment is greater than the product of the threshold coefficient and the truncation standard deviation, it is determined to be an outlier, and the most recent effective increment value is used to replace the outlier.

7. The method for dynamic compensation of wear error in a tilting centerless grinder according to claim 1, characterized in that, The process of applying exponentially damped accumulation to the dose factor increment after trimming to generate asymptotic softening coefficients includes: The current trimmed dose factor increment after smoothing is normalized to the current conventional wear baseline, and then the normalized increment is superimposed on the softening asymptotic coefficient of the previous workpiece to obtain the softening asymptotic coefficient of the current workpiece. The initial softening degree progressive coefficient is zero, and as the conventional wear reference amount increases, the growth rate of the softening degree progressive coefficient caused by the same increment decreases.

8. The method for dynamic compensation of wear error in a tilting centerless grinder according to claim 1, characterized in that, The process of generating a dynamic wear offset based on the softening degree progressive coefficient and a preset conventional wear reference value, through saturation limiting mapping, includes: The saturation limiting mapping calculation uses a hyperbolic tangent function. Input variables are generated based on the softening degree asymptotic coefficient and the conventional wear reference amount. The output value of the hyperbolic tangent function is multiplied by a preset saturation upper limit to obtain the dynamic wear offset. The saturation upper limit is determined according to the dimensional tolerance of the workpiece. When the softening degree asymptotic coefficient increases, the dynamic wear offset approaches and does not exceed the saturation upper limit at a decreasing rate.

9. A dynamic wear error compensation system for an inclined centerless grinder, characterized in that, include: A processor and a memory, the memory storing computer program instructions that, when executed by the processor, implement a dynamic wear error compensation method for a tilting centerless grinder according to any one of claims 1-8.

Images