Rotating shaft machining control method based on multi-angle monitoring technology

Abstract

The invention belongs to the technical field of machining control, and discloses a rotating shaft machining control method based on a multi-angle monitoring technology, and the method comprises the steps: constructing a monitoring point deviation prediction model, and predicting the radial cutter relieving deviation of each monitoring point according to the attributes of a bar and machining parameters; before batch processing, generating a predicted deviation curve and initializing a model correction factor; a residual compensation curve is initialized before single-piece machining, the predicted deviation curve is corrected through a model correction factor and then summed with the residual compensation curve to obtain a total to-be-compensated curve, machining parameters are corrected according to the total to-be-compensated curve, and a machine tool is controlled to execute machining; after each round of machining is completed, collecting measurement data, calculating an actual error curve, and updating a model correction factor and a residual compensation curve according to the actual error curve; axial differential compensation is achieved through a predicted deviation curve, and the problem that a fixed compensation value cannot adapt to non-uniform distribution of cutter relieving deviation is solved; and self-adaptive adjustment of working condition changes is achieved through a dynamic updating mechanism, and the stability of machining precision is guaranteed.

Description

Technical Field

[0001] This invention relates to the field of machining control technology, and more specifically, to a shaft machining control method based on multi-angle monitoring technology. Background Technology

[0002] Shafts are fundamental components widely used in mechanical manufacturing, playing a crucial role in supporting rotating parts in structures such as brackets, hinges, connectors, and transmission mechanisms. Because support shafts need to provide stable support for rotating parts within a limited space, their structural design typically prioritizes lightweight and compactness; therefore, these shafts often feature a slender shape with a small diameter and a large length-to-diameter ratio. With the increasing demands for product performance and reliability in modern manufacturing, the machining accuracy of shafts has become a key factor affecting assembly quality.

[0003] During turning, the workpiece undergoes elastic deformation due to the cutting force, causing the tool's actual cutting position relative to the workpiece to deviate from its theoretical position. This phenomenon is called tool deflection. Tool deflection is particularly pronounced in the machining of slender shafts. When the workpiece is clamped in a cantilevered state, the rigidity decreases the further away from the clamping end, resulting in greater flexural deformation under the cutting force. Tool deflection is one of the main reasons for the high scrap rate and low production efficiency in the machining of shaft-type parts.

[0004] Currently, the control methods for tool deflection deviation mainly rely on manual compensation based on the operator's experience. Technicians pre-set a fixed compensation value based on factors such as workpiece material, tool parameters, and cutting parameters, and then adjust the cutting depth or tool position during machining to counteract the tool deflection effect. However, this fixed compensation value method has significant limitations. First, a fixed compensation value cannot adapt to the non-uniform distribution of tool deflection deviation along the axial direction. Due to differences in rigidity at different locations on the workpiece, the amount of tool deflection varies at each axial position; the amount of tool deflection is smaller near the clamping end and larger further away. Using a single fixed compensation value for global compensation inevitably leads to over-compensation at some positions and under-compensation at others, making precise control of each position impossible and affecting the overall dimensional accuracy and cylindricity of the workpiece. Second, a fixed compensation value cannot adapt to the dynamic changes in working conditions during machining. As machining continues, the tool gradually wears down, the geometry of the cutting edge changes, leading to changes in cutting force characteristics and corresponding changes in tool deflection deviation. Furthermore, factors such as machine tool thermal deformation and fluctuations in bar stock material also affect tool deflection deviation. Fixed compensation values ​​may be relatively accurate in the early stages of processing, but as the above factors accumulate and change, the compensation effect will gradually deviate from the optimal state, leading to a continuous decline in processing accuracy. Summary of the Invention

[0005] To overcome the aforementioned problems in the prior art, this invention proposes a shaft machining control method based on multi-angle monitoring technology to solve the above problems.

[0006] This invention provides the following technical solution: A method for controlling the machining of a rotating shaft based on multi-angle monitoring technology, comprising: A monitoring point deviation prediction model is constructed to predict the basic radial deviation of at least three monitoring points based on bar stock properties and processing parameters, wherein the monitoring points are determined according to the processing parameters; Before batch processing, the preset processing parameters for each round of workpiece are obtained, and the corresponding bar stock properties are input into the monitoring point deviation prediction model to obtain the expected deviation at each monitoring point for each round, and generate the expected deviation curve for each round; a model correction factor is initialized. Before processing a single piece, a residual compensation curve is initialized. For the current processing round, the expected deviation curve is corrected using the model correction factor. After correction, the curve is summed with the residual compensation curve to obtain the total curve to be compensated. The machining parameters are corrected based on the total compensation curve to obtain the corrected machining parameters, and the machine tool is controlled to perform the current round of machining. After each round of processing is completed, the measurement data of each monitoring point is collected to calculate the actual error curve; the model correction factor and residual compensation curve are updated based on the actual error curve.

[0007] Preferably, the bar stock attributes include the bar stock diameter and machining length; the machining parameters include the depth of cut curve, feed rate and spindle speed; wherein the depth of cut curve is established based on a global coordinate system, the global coordinate system having the intersection of the machine tool spindle axis and the bar stock clamping section as the origin, the spindle axis as the X-axis, and the radial direction perpendicular to the spindle axis as the Y-axis; The steps for determining the monitoring point include: extracting the starting coordinates and ending coordinates from the cutting depth curve; and determining the starting coordinates, ending coordinates, and the midpoint coordinates between them as the monitoring point.

[0008] Preferably, the steps for constructing the monitoring point deviation prediction model include: Acquire sample data, which includes processing parameters, bar stock properties, and measured radial deviations at each monitoring point after cutting. Input feature vectors are extracted from the processing parameters and bar stock properties in each set of sample data; the measured radial deviations of each monitoring point in each set of sample data are extracted and combined into a multidimensional target label vector; An initial model is constructed using a machine learning regression algorithm that supports multiple outputs. The input feature vector is input into the model to obtain the predicted output vector. The loss function between the predicted output vector and the target label vector is calculated. The internal weight parameters of the model are adjusted through an iterative algorithm to minimize the loss function until the model converges, thus obtaining the monitoring point deviation prediction model.

[0009] Preferably, the step of generating the expected deviation curve for each round includes: For each round, the X-axis coordinates of each monitoring point in the global coordinate system are obtained and the corresponding expected deviation is used as the Y-axis coordinate to obtain a two-dimensional coordinate point set; A polynomial fitting algorithm is used to fit the two-dimensional coordinate point set to generate the expected deviation curve corresponding to this round.

[0010] Preferably, the model correction factor includes a gain coefficient and a bias coefficient, and the model correction factor is initialized by setting the gain coefficient to 1 and the bias coefficient to 0. The step of correcting the predicted deviation curve using a model correction factor includes: Extract several discrete data points on the expected deviation curve; multiply the Y coordinate value of each discrete data point by the gain coefficient and add the bias coefficient to calculate the corresponding corrected Y coordinate value; Using the corrected Y-coordinate value and its corresponding X-coordinate value, a new curve is fitted to complete the correction of the expected deviation curve.

[0011] Preferably, the step of correcting the processing parameters based on the total compensation curve to obtain the corrected processing parameters includes: Obtain the X-axis interval where the Y-axis coordinate is greater than 0 on the cutting depth curve, and sum the total curve to be compensated and the cutting depth curve within the interval to obtain the corrected cutting depth curve; Based on the relative change in the depth of cut curve, the feed rate and spindle speed of this cycle are adjusted in accordance with the preset process matching rules; The modified cutting depth curve, the adjusted feed rate, and the spindle speed are combined to form the modified machining parameters.

[0012] Preferably, the step of collecting measurement data from each monitoring point and calculating the actual error curve includes: The actual error curve obtained from the previous round of calculation is used as the base state curve. If the current round is the first round, a zero-value curve covering the range of bar stock processing length is initialized as the base state curve. Calculate the corresponding error based on the measurement data of each monitoring point, obtain the X-axis coordinate of each monitoring point in the global coordinate system and use the corresponding error as the Y-axis coordinate to obtain a two-dimensional coordinate point set; The two-dimensional coordinate point set is fitted using a polynomial fitting algorithm to generate a local error curve segment that covers the minimum range of the X-axis coordinates of each monitoring point. The interval in the basic state curve that coincides with the X-axis coordinate range of the local error curve segment is replaced with the local error curve segment to generate a new actual error curve.

[0013] Preferably, the update step of the model correction factor includes: The total deviation curve is obtained by summing the actual error curve and the expected deviation curve. By selecting characteristic positions on the curve and comparing the magnitude and offset relationship between the total deviation curve and the expected deviation curve, the gain coefficient and bias coefficient are calculated using linear regression to update the model correction factor.

[0014] Preferably, the method for updating the residual compensation curve includes updating the actual error curve to a new residual compensation curve.

[0015] This invention provides a shaft machining control method based on multi-angle monitoring technology, which has the following beneficial effects: By constructing a monitoring point deviation prediction model, the distribution of tool deflection deviation at each axial position can be predicted based on the bar stock properties and machining parameters. This generates a predicted deviation curve, which is then used to differentiate the depth of cut. Compared to the traditional fixed compensation value method, this approach fully considers the non-uniform distribution of tool deflection deviation along the axial direction, achieving precise control of each axial position and effectively improving the overall dimensional accuracy and cylindricity of the workpiece.

[0016] By setting a model calibration factor and a residual compensation curve, a dual-channel control mechanism combining predictive compensation and feedback correction was constructed. The model calibration factor is dynamically adjusted based on the comparison between measured and predicted data, correcting model prediction deviations caused by factors such as tool wear, machine tool thermal deformation, and fluctuations in bar stock material, allowing the compensation strategy to continuously adapt to changes in machining conditions. The residual compensation curve records and transmits the residual error from each round, which is gradually reduced through subsequent rounds of machining, achieving cross-round error accumulation control. The combined efforts of these two mechanisms enable the system to adaptively adjust, avoiding the accuracy degradation problem caused by the fixed compensation value method's inability to adapt to dynamic changes, and ensuring the stability and consistency of machining accuracy during batch processing. Attached Figure Description

[0017] Figure 1 This is a flowchart illustrating a shaft machining control method based on multi-angle monitoring technology according to the present invention. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention. Example

[0019] Please see Figure 1 In this embodiment, a shaft machining control method based on multi-angle monitoring technology includes: S1. Construct a monitoring point deviation prediction model for predicting the basic radial deviation of at least three monitoring points based on bar stock properties and processing parameters, wherein the monitoring points are determined according to the processing parameters; The bar stock attributes include the bar stock diameter and machining length; the machining parameters include the depth of cut curve, feed rate and spindle speed; wherein the depth of cut curve is established based on a global coordinate system, the global coordinate system taking the intersection of the machine tool spindle axis and the bar stock clamping section as the origin, the spindle axis as the X-axis, and the radial direction perpendicular to the spindle axis as the Y-axis; The steps for determining the monitoring point include: extracting the starting coordinates and ending coordinates from the cutting depth curve; and determining the starting coordinates, ending coordinates, and the midpoint coordinates between them as the monitoring point.

[0020] The steps for constructing the monitoring point deviation prediction model include: Acquire sample data, which includes processing parameters, bar stock properties, and measured radial deviations at each monitoring point after cutting. Input feature vectors are extracted from the processing parameters and bar stock properties in each set of sample data; the measured radial deviations of each monitoring point in each set of sample data are extracted and combined into a multidimensional target label vector; An initial model is constructed using a machine learning regression algorithm that supports multiple outputs. The input feature vector is input into the model to obtain the predicted output vector. The loss function between the predicted output vector and the target label vector is calculated. The internal weight parameters of the model are adjusted through an iterative algorithm to minimize the loss function until the model converges, thus obtaining the monitoring point deviation prediction model.

[0021] In this embodiment, it should be noted that the monitoring point deviation prediction model is used to predict radial tool deflection deviation caused by factors such as cutting force, tool deformation, and workpiece deflection during the machining of the shaft. When turning slender shaft-like workpieces, due to the poor rigidity of the workpiece and its cantilevered state, elastic deformation will occur during the cutting process, resulting in the actual cutting depth being less than the set value. This phenomenon is called tool deflection.

[0022] It should be noted that setting the number of monitoring points to at least three is based on the mathematical requirement of curve fitting. Since the tool deviation exhibits a non-linear distribution along the axial direction, the deviation is typically smaller near the clamping end and larger further away. Using the starting coordinates, ending coordinates, and the midpoint between them as monitoring points can capture the basic shape of the deviation curve. Three points can support quadratic polynomial fitting, effectively reflecting the trend of deviation changes. In practical applications, the number of monitoring points can be increased according to machining accuracy requirements, for example, setting five or more monitoring points to improve fitting accuracy.

[0023] It should be noted that the establishment of the global coordinate system provides a unified spatial reference for the entire machining control process. The origin is set at the intersection of the machine tool spindle axis and the bar stock clamping section, ensuring the correspondence between the coordinate system and the actual machining state. The X-axis extends along the spindle axis, representing the axial position; a larger X-value indicates a greater distance from the clamping end. The Y-axis represents the radial deviation; a positive value indicates that the actual workpiece diameter is greater than the target diameter, and a negative value indicates that the actual diameter is less than the target diameter. The depth-of-cut curve defines the preset depth of cut for each axial position within this coordinate system, facilitating overlay calculations with the deviation curve and compensation curve.

[0024] It's important to note that the multi-dimensional target label vector setting enables the model to simultaneously predict the deviation values ​​of multiple monitoring points. Taking three monitoring points as an example, the target label vector is a three-dimensional vector, corresponding to the radial deviations of the starting position, midpoint position, and ending position, respectively. This multi-output design, compared to building a separate model for each monitoring point, can better learn the intrinsic correlation between deviations at different positions, improving the overall consistency of predictions. Machine learning regression algorithms that support multiple outputs include, but are not limited to, multi-output random forest regression, multi-output gradient boosting regression, and multi-layer neural network regression. These algorithms can accept multi-dimensional feature inputs and generate multi-dimensional prediction outputs. The loss function typically uses mean squared error or mean absolute error to calculate the deviation between the predicted and measured values. Iterative algorithms, such as gradient descent, gradually reduce the loss function by repeatedly adjusting the model parameters. The criteria for model convergence can be conditions such as the loss function value falling below a set threshold, the change in the loss function being less than a set value after several consecutive iterations, or reaching the maximum number of iterations.

[0025] S2. Before batch processing, obtain the preset processing parameters for each round of workpiece, combine them with the corresponding bar stock properties to input the monitoring point deviation prediction model, obtain the expected deviation of each round at each monitoring point, and generate the expected deviation curve for each round; initialize a model correction factor. The step of generating the expected deviation curve for each round includes: For each round, the X-axis coordinates of each monitoring point in the global coordinate system are obtained and the corresponding expected deviation is used as the Y-axis coordinate to obtain a two-dimensional coordinate point set; A polynomial fitting algorithm is used to fit the two-dimensional coordinate point set to generate the expected deviation curve corresponding to this round.

[0026] In this embodiment, it should be noted that batch processing refers to a production mode in which multiple workpieces of the same specification are processed sequentially using the same procedures. The processing of each workpiece typically includes multiple rounds, such as roughing, semi-finishing, and finishing. The processing parameters for each round are preset according to the process requirements. Before batch processing begins, the system pre-acquires the processing parameters for all rounds and, in conjunction with the current bar stock properties, calculates the expected deviation for each round using a monitoring point deviation prediction model, providing basic data for subsequent real-time compensation control.

[0027] It should be noted that the bar stock properties change dynamically with each machining pass. After each round of cutting, the bar stock diameter decreases accordingly; therefore, the bar stock properties for each pass need to be calculated based on the cutting depth of the previous pass. The predicted deviation curve describes the distribution of tool deviation along the workpiece axis. Since the monitoring point deviation prediction model only outputs the predicted deviation values ​​at discrete monitoring points, curve fitting is needed to extend these discrete points into a continuous curve to estimate the deviation at any axial position. The polynomial fitting algorithm determines the polynomial coefficients to make the fitted curve approximate the known data points as closely as possible. For the case of three monitoring points, a quadratic polynomial is typically used for fitting; the fitted curve is usually a quadratic function, solved using the least squares method.

[0028] It should be noted that the model correction factor is used to correct systematic biases in the monitoring point deviation prediction model. Since the prediction model is trained based on historical sample data, when actual processing conditions differ from the training data, the model's predicted values ​​may be generally too high or too low. The model correction factor uses two parameters, the gain coefficient and the bias coefficient, to linearly transform the predicted deviation curve, making the corrected predicted values ​​closer to the actual tool deviation. Initially, the gain coefficient is set to 1 and the bias coefficient to 0, indicating that the model's original prediction output is fully trusted in the initial state. The correction factor is then gradually adjusted based on actual processing feedback.

[0029] S3. Before processing a single piece, initialize a residual compensation curve. For the current processing round, use the model correction factor to correct the expected deviation curve. After correction, sum the curve with the residual compensation curve to obtain the total curve to be compensated. The model correction factor includes a gain coefficient and a bias coefficient. The model correction factor is initialized by setting the gain coefficient to 1 and the bias coefficient to 0. The step of correcting the predicted deviation curve using a model correction factor includes: Extract several discrete data points on the expected deviation curve; multiply the Y coordinate value of each discrete data point by the gain coefficient and add the bias coefficient to calculate the corresponding corrected Y coordinate value; Using the corrected Y-coordinate value and its corresponding X-coordinate value, a new curve is fitted to complete the correction of the expected deviation curve.

[0030] In this embodiment, it should be noted that single-piece machining refers to the complete machining process of a single workpiece in a batch, including multiple rounds from roughing to finishing. Before machining each workpiece, the residual compensation curve needs to be initialized to ensure that the machining control of each workpiece is independent of each other, and the machining state of the previous workpiece will not affect the compensation strategy of the next workpiece. The residual compensation curve is initialized as a zero-value curve covering the machining length range of the bar stock, indicating that there are no historical errors that need to be compensated.

[0031] It should be noted that the total compensation curve is obtained by summing the corrected predicted deviation curve and the residual compensation curve. The corrected predicted deviation curve reflects the model's prediction of the tool deflection deviation in the current round and is used to actively compensate for the expected tool deflection phenomenon; the residual compensation curve reflects the machining errors left over from previous rounds and is used to passively correct the deviations that have already occurred. The sum of the two forms the total compensation curve, achieving comprehensive control of machining errors.

[0032] It should be noted that initializing the gain coefficient to 1 indicates that the magnitude of the prediction deviation is not scaled, and initializing the bias coefficient to 0 indicates that no additional fixed offset is applied. In other words, the original output of the monitoring point deviation prediction model is completely trusted in the initial state. As the machining process progresses, the system gradually adjusts the correction factor based on actual measurement feedback, so that the corrected prediction value gradually approaches the actual tool deviation.

[0033] S4. Correct the machining parameters according to the total compensation curve to obtain the corrected machining parameters, and control the machine tool to execute the current round of machining; The step of obtaining the corrected processing parameters by correcting the processing parameters based on the total compensation curve includes: Obtain the X-axis interval where the Y-axis coordinate is greater than 0 on the cutting depth curve, and sum the total curve to be compensated and the cutting depth curve within the interval to obtain the corrected cutting depth curve; Based on the relative change in the depth of cut curve, the feed rate and spindle speed of this cycle are adjusted in accordance with the preset process matching rules; The modified cutting depth curve, the adjusted feed rate, and the spindle speed are combined to form the modified machining parameters.

[0034] In this embodiment, it should be noted that the X-axis interval where the Y-axis coordinate is greater than 0 on the depth of cut curve represents the effective machining area of ​​the current cycle. Within this interval, the tool and workpiece make actual cutting contact, resulting in tool deflection deviation; outside this interval, the tool does not participate in cutting, and therefore no compensation is performed. The total compensation curve is only superimposed on the depth of cut curve within the effective machining area to retain the compensation amount for the unmachined area in this cycle.

[0035] It should be noted that the corrected cutting depth curve is obtained by summing the total compensation curve with the original cutting depth curve. Since tool deflection can cause the actual cutting depth to be less than the set value, this deviation is compensated by increasing the set cutting depth. For example, if the preset cutting depth at a certain position is 0.5mm and the total compensation amount is 0.08mm, then the corrected cutting depth is 0.58mm, allowing the actual cutting depth to approach the target value of 0.5mm under the influence of tool deflection.

[0036] It should be noted that changes in the depth of cut affect the magnitude of the cutting force, which in turn affects machining stability and surface quality. When the corrected depth of cut increases relative to the original value, the cutting force increases accordingly. To ensure the stability of the machining process, it is necessary to appropriately reduce the feed rate or adjust the spindle speed. Conversely, when the depth of cut decreases, the feed rate can be appropriately increased to improve machining efficiency. The process matching rules are preset based on factors such as machine tool characteristics, tool parameters, and workpiece material, and are used to determine the correspondence between the change in depth of cut and the adjustment of feed rate and spindle speed.

[0037] It should be noted that the corrected machining parameters include three parts: the corrected depth-of-cut curve, the adjusted feed rate, and the spindle speed. The machine tool CNC system generates the tool path and spindle control commands based on the corrected machining parameters and executes the current machining cycle. The corrected depth-of-cut curve is transmitted to the CNC system in the form of discrete data points or a parametric curve. The CNC system converts this into a continuous toolpath through interpolation calculations, achieving precise control of the depth of cut at each axial position.

[0038] S5. After each round of processing is completed, collect the measurement data of each monitoring point to calculate the actual error curve; update the model correction factor and residual compensation curve based on the actual error curve.

[0039] The steps for collecting measurement data from each monitoring point and calculating the actual error curve include: The actual error curve obtained from the previous round of calculation is used as the base state curve. If the current round is the first round, a zero-value curve covering the range of bar stock processing length is initialized as the base state curve. Calculate the corresponding error based on the measurement data of each monitoring point, obtain the X-axis coordinate of each monitoring point in the global coordinate system and use the corresponding error as the Y-axis coordinate to obtain a two-dimensional coordinate point set; The two-dimensional coordinate point set is fitted using a polynomial fitting algorithm to generate a local error curve segment that covers the minimum range of the X-axis coordinates of each monitoring point. The interval in the basic state curve that coincides with the X-axis coordinate range of the local error curve segment is replaced with the local error curve segment to generate a new actual error curve.

[0040] The update steps for the model correction factor include: The total deviation curve is obtained by summing the actual error curve and the expected deviation curve. By selecting characteristic positions on the curve and comparing the magnitude and offset relationship between the total deviation curve and the expected deviation curve, the gain coefficient and bias coefficient are calculated using linear regression to update the model correction factor.

[0041] The method for updating the residual compensation curve includes updating the actual error curve to a new residual compensation curve.

[0042] In this embodiment, it should be noted that online measurement of the workpiece after each round of processing is a key step in achieving closed-loop control. Measurement data can be acquired through an online measurement system equipped on the machine tool, such as a contact probe or laser measuring device, to measure the actual diameter at each monitoring point without disassembling the workpiece. The error is calculated as the difference between the actual measured diameter and the target diameter; a positive value indicates that the workpiece diameter is too large, meaning insufficient material removal, while a negative value indicates that the workpiece diameter is too small, meaning excessive material removal.

[0043] It should be noted that the actual error curve is maintained using a local update method to fully record the machining status of each area of ​​the workpiece. Since the machining areas may differ between different rounds—for example, a round might only cut a portion of the workpiece—only the curve segment corresponding to the effective machining area of ​​that round is updated, while historical error records for other areas are retained. The base state curve inherits the error distribution information from previous rounds, and the measurement results of the current round are incorporated into it through local replacement, forming the actual error curve that reflects the overall error status of the workpiece.

[0044] It should be noted that the model correction factor update is based on a comparative analysis of the predicted deviation and the actual deviation. The total deviation curve is obtained by summing the actual error curve and the predicted deviation curve, and its physical meaning is the actual tool deflection deviation that occurred in this round of machining. Comparing the total deviation curve with the predicted deviation curve, if they are consistent, it indicates that the model prediction is accurate, and the correction factor should approach the initial state where the gain coefficient is 1 and the bias coefficient is 0; if the amplitude of the total deviation curve is greater than that of the predicted deviation curve, it indicates that the model prediction is too small, and the gain coefficient needs to be increased; conversely, the gain coefficient needs to be decreased. The linear regression method calculates the optimal gain coefficient and bias coefficient by minimizing the fitting error between the values ​​at the characteristic positions of the two curves.

[0045] It should be noted that the residual compensation curve is updated using a direct replacement method, taking the current actual error curve as the residual compensation curve for the next round. This design is based on the principle of error propagation: if an error still exists after the current round of processing, this error will be included in the compensation calculation as the initial state for the next round, and eliminated through subsequent rounds of processing, thereby improving processing accuracy.

[0046] In the several embodiments provided by this invention, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only one method, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0047] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention.

[0048] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for controlling the machining of a rotating shaft based on multi-angle monitoring technology, characterized in that, include: A monitoring point deviation prediction model is constructed to predict the basic radial deviation of at least three monitoring points based on bar stock properties and processing parameters, wherein the monitoring points are determined according to the processing parameters; Before batch processing, the preset processing parameters for each round of workpiece are obtained, and the corresponding bar stock properties are input into the monitoring point deviation prediction model to obtain the expected deviation at each monitoring point for each round, and generate the expected deviation curve for each round; a model correction factor is initialized. Before processing a single piece, a residual compensation curve is initialized. For the current processing round, the expected deviation curve is corrected using the model correction factor. After correction, the curve is summed with the residual compensation curve to obtain the total curve to be compensated. The machining parameters are corrected based on the total compensation curve to obtain the corrected machining parameters, and the machine tool is controlled to perform the current round of machining. After each round of processing is completed, the measurement data of each monitoring point is collected to calculate the actual error curve; the model correction factor and residual compensation curve are updated based on the actual error curve.

2. The shaft machining control method based on multi-angle monitoring technology according to claim 1, characterized in that, The bar stock attributes include the bar stock diameter and machining length; the machining parameters include the depth of cut curve, feed rate and spindle speed; wherein the depth of cut curve is established based on a global coordinate system, the global coordinate system taking the intersection of the machine tool spindle axis and the bar stock clamping section as the origin, the spindle axis as the X-axis, and the radial direction perpendicular to the spindle axis as the Y-axis; The steps for determining the monitoring point include: extracting the starting coordinates and ending coordinates from the cutting depth curve; and determining the starting coordinates, ending coordinates, and the midpoint coordinates between them as the monitoring point.

3. The shaft machining control method based on multi-angle monitoring technology according to claim 2, characterized in that, The steps for constructing the monitoring point deviation prediction model include: Acquire sample data, which includes processing parameters, bar stock properties, and measured radial deviations at each monitoring point after cutting. Input feature vectors are extracted from the processing parameters and bar stock properties in each set of sample data; the measured radial deviations of each monitoring point in each set of sample data are extracted and combined into a multidimensional target label vector; An initial model is constructed using a machine learning regression algorithm that supports multiple outputs. The input feature vector is input into the model to obtain the predicted output vector. The loss function between the predicted output vector and the target label vector is calculated. The internal weight parameters of the model are adjusted through an iterative algorithm to minimize the loss function until the model converges, thus obtaining the monitoring point deviation prediction model.

4. The shaft machining control method based on multi-angle monitoring technology according to claim 2, characterized in that, The step of generating the expected deviation curve for each round includes: For each round, the X-axis coordinates of each monitoring point in the global coordinate system are obtained and the corresponding expected deviation is used as the Y-axis coordinate to obtain a two-dimensional coordinate point set; A polynomial fitting algorithm is used to fit the two-dimensional coordinate point set to generate the expected deviation curve corresponding to this round.

5. The shaft machining control method based on multi-angle monitoring technology according to claim 4, characterized in that, The model correction factor includes a gain coefficient and a bias coefficient. The model correction factor is initialized by setting the gain coefficient to 1 and the bias coefficient to 0. The step of correcting the predicted deviation curve using a model correction factor includes: Extract several discrete data points on the expected deviation curve; multiply the Y coordinate value of each discrete data point by the gain coefficient and add the bias coefficient to calculate the corresponding corrected Y coordinate value; Using the corrected Y-coordinate value and its corresponding X-coordinate value, a new curve is fitted to complete the correction of the expected deviation curve.

6. The shaft machining control method based on multi-angle monitoring technology according to claim 5, characterized in that, The step of obtaining the corrected processing parameters by correcting the processing parameters based on the total compensation curve includes: Obtain the X-axis interval where the Y-axis coordinate is greater than 0 on the cutting depth curve, and sum the total curve to be compensated and the cutting depth curve within the interval to obtain the corrected cutting depth curve; Based on the relative change in the depth of cut curve, the feed rate and spindle speed of this cycle are adjusted in accordance with the preset process matching rules; The modified cutting depth curve, the adjusted feed rate, and the spindle speed are combined to form the modified machining parameters.

7. The shaft machining control method based on multi-angle monitoring technology according to claim 6, characterized in that, The steps for collecting measurement data from each monitoring point and calculating the actual error curve include: The actual error curve obtained from the previous round of calculation is used as the base state curve. If the current round is the first round, a zero-value curve covering the range of bar stock processing length is initialized as the base state curve. Calculate the corresponding error based on the measurement data of each monitoring point, obtain the X-axis coordinate of each monitoring point in the global coordinate system and use the corresponding error as the Y-axis coordinate to obtain a two-dimensional coordinate point set; The two-dimensional coordinate point set is fitted using a polynomial fitting algorithm to generate a local error curve segment that covers the minimum range of the X-axis coordinates of each monitoring point. The interval in the basic state curve that coincides with the X-axis coordinate range of the local error curve segment is replaced with the local error curve segment to generate a new actual error curve.

8. The shaft machining control method based on multi-angle monitoring technology according to claim 7, characterized in that, The update steps for the model correction factor include: The total deviation curve is obtained by summing the actual error curve and the expected deviation curve. By selecting characteristic positions on the curve and comparing the magnitude and offset relationship between the total deviation curve and the expected deviation curve, the gain coefficient and bias coefficient are calculated using linear regression to update the model correction factor.

9. A shaft machining control method based on multi-angle monitoring technology according to claim 8, characterized in that, The method for updating the residual compensation curve includes updating the actual error curve to a new residual compensation curve.

Images