A Modeling Method for Machine Tool Thermal Error Based on Weight Update of Thermal Parameters under Variable Operating Conditions

Abstract

A machine tool thermal error modeling method based on weighted updates of thermal parameters under varying operating conditions is proposed. This method divides machine tool thermal errors into sources and regions: spindle-side, table-side, feed axis-side, and machining area thermal errors. Operating conditions are categorized into internal heat source conditions and external boundary conditions. The method establishes mappings between primary operating conditions and equivalent heat source intensity, secondary operating conditions, thermal path coupling relationships, and relationships between external boundary conditions and equivalent disturbance intensity. By comprehensively identifying regional thermal inertia coefficients and initial weights, a local thermal error model for the non-machining area is established and mapped to the tool and workpiece ends through kinematic chain coordinate transformation. The difference between the measured total thermal error of cutting and the mapped thermal error of the non-machining area is used as the equivalent residual for the machining area. A neural network is used to update the equivalent heat source intensity weights, thermal path coupling weights, and external boundary equivalent disturbance weights in real time based on varying operating condition information. This method solves the problem of insufficient generalization ability of thermal error models under varying operating conditions, achieving high-precision dynamic modeling and prediction of machine tool thermal errors.

Description

Technical Field

[0001] This invention belongs to the field of CNC machine tool precision control technology, specifically a machine tool thermal error modeling method based on updating the weights of thermal parameters under varying operating conditions. This method dynamically updates the weights of thermal parameters using a neural network to adapt to varying operating conditions, and can be widely applied to thermal error prediction and real-time compensation for various CNC machine tools under complex machining conditions. Background Technology

[0002] As core equipment in modern manufacturing, CNC machine tools directly determine the quality and performance of products through their machining accuracy. Among the many factors affecting the machining accuracy of CNC machine tools, thermal error is one of the most critical. Studies have shown that in precision machining, thermal errors caused by machine tool thermal deformation can account for 40% to 70% of the total machining error, and this proportion will further increase with the improvement of machine tool accuracy. During operation, internal heat sources such as the spindle motor, feed motor, bearing friction, and lead screw and nut pairs generate a large amount of heat. Simultaneously, they are affected by external boundary conditions such as ambient temperature fluctuations and coolant temperature changes, leading to non-uniform temperature fields and thermoelastic deformation in various machine tool components. This alters the ideal relative position between the tool and the workpiece, ultimately resulting in machining thermal errors. Therefore, establishing an accurate and robust thermal error prediction model and performing real-time compensation is a key approach to improving the machining accuracy of CNC machine tools.

[0003] Currently, scholars both domestically and internationally have conducted extensive research on machine tool thermal error modeling and compensation technology, resulting in various technical approaches. The first category is modeling methods based on physical mechanisms. These methods, based on fundamental principles of heat transfer, describe the physical relationship between temperature field and thermal deformation by constructing heat conduction equations, thermal network models, or finite element models of key machine tool components, offering clear physical interpretability. For example, a heat transfer model of an electric spindle system can be established based on the thermal resistance network method. By solving the heat balance equation, the temperature distribution of each thermal node can be obtained, and then the amount of thermal deformation can be calculated based on the principles of thermoelasticity. However, these methods require precise setting of numerous geometric parameters, material thermal properties, contact thermal resistance, and complex boundary conditions. In practical engineering applications, parameter acquisition is difficult, and when facing unsteady heat transfer processes under varying operating conditions, it is difficult to accurately describe the dynamic effects of changes in heat source intensity, alterations in heat transfer paths, and fluctuations in external boundary conditions on the temperature field and thermal deformation, often making it difficult to guarantee model accuracy.

[0004] The second category is data-driven modeling methods. These methods collect temperature data from key parts of the machine tool using temperature sensors and simultaneously measure thermal errors using displacement sensors. Algorithms such as multiple linear regression, support vector regression, artificial neural networks, and deep learning are then used to directly establish the mapping relationship between temperature and thermal error. These methods do not require a deep understanding of the physical mechanisms of thermal error generation, have strong fitting capabilities for nonlinear relationships, and can achieve high prediction accuracy under specific operating conditions. For example, long short-term memory networks can be used to establish a dynamic mapping relationship between temperature time series and thermal error, convolutional neural networks can be used to process infrared thermograms to predict spindle thermal deformation, or parallel deep belief networks can be used to achieve simultaneous prediction of multi-directional thermal errors. However, purely data-driven methods are essentially "black box" models, lacking physical interpretability. The model parameters lack clear physical meaning, and when machining conditions change (such as spindle speed switching, feed rate adjustment, and ambient temperature fluctuations), the mapping relationship in the model cannot adaptively adjust, resulting in severely insufficient generalization ability. Furthermore, data-driven methods heavily rely on large amounts of high-quality training data. In scenarios with varying operating conditions and across different equipment, data acquisition costs are high and it is difficult to cover all operating conditions.

[0005] The third category is the hybrid modeling approach driven by both mechanistic and data-driven methods. This type of approach attempts to combine the interpretability of physical mechanism models with the fitting ability of data-driven models. Existing hybrid-driven schemes include: calculating the temperature field distribution or thermal error baseline value through a physical model, and then using a machine learning model to compensate for the prediction residuals of the physical model, in order to correct the bias caused by parameter uncertainty or simplification assumptions; or using both theoretical and measured temperature values ​​calculated by the physical model as input features of the data-driven model to improve the completeness of the model input information; or using prior knowledge provided by the physical model to guide the training process of the data-driven model, reducing the dependence on training data. However, existing hybrid-driven methods still have significant technical limitations.

[0006] First, existing hybrid-driven methods generally treat the physical model as a fixed "computational module." Once established, its internal parameters are no longer adjusted, and corrections are only made at the output or residual ends through data-driven modeling. When the processing conditions change significantly, the core parameters on which the physical model depends, such as the heat source intensity, heat transfer path weights, and external boundary conditions, have actually changed fundamentally. The fixed-parameter physical model cannot reflect this change, leading to a systematic deviation in the "benchmark" of the physical model, and the correction capability of the data-driven model reaches its limit.

[0007] Secondly, existing technologies typically model the entire machine tool as a single thermal system when establishing physical models, or perform thermal analysis only on a single component (such as the spindle or feed system). This fails to fully consider the different influence mechanisms and contributions of various thermal error sources (spindle side, table side, feed axis side, and machining area) on the final machining error. In actual machining, the heat source characteristics and heat transfer paths of different areas differ fundamentally: spindle-side thermal error is mainly affected by the heat generated by the spindle motor and bearing friction; table-side thermal error is mainly affected by the heat generated by the table transmission system; feed axis-side thermal error is mainly affected by the frictional heat of the lead screw and nut pair and the guide rail; while machining area thermal error is directly and strongly influenced by cutting heat, the contact state between the tool and the workpiece, and the cooling effect of the cutting fluid. If these different sources of thermal error are treated as a single entity and a uniform temperature-thermal error mapping relationship is established, the model will struggle to accurately distinguish the differentiated mechanisms of different heat sources, leading to decreased prediction accuracy and generalization ability.

[0008] Furthermore, existing hybrid drive methods do not address the independent identification of "machining area thermal error." During the cutting process, the cutting heat, frictional heat, and dynamic cooling effect of the cutting fluid in the machining area (tool-workpiece contact interface) generate additional thermal errors. These thermal errors are difficult to measure or calculate directly using temperature sensors mounted on the machine tool structure. Existing technologies typically use the measured total thermal error under cutting conditions directly as the model training label, or simply treat cutting heat as an additional heat source, failing to effectively distinguish between the two different types of error sources: "non-machining area thermal error caused by internal machine tool heat sources" and "machining area thermal error caused by the cutting process." This compromises the physical consistency of the model.

[0009] Finally, existing technologies lack effective mechanisms for the dynamic updating of model parameters. Some studies have attempted to use classic adaptive algorithms such as Kalman filtering and recursive least squares to correct model parameters online. However, these methods can usually only perform linear recursive corrections on a single scalar parameter or a few parameters. When faced with multiple thermal parameters (heat source intensity weight, thermal path coupling weight, external boundary disturbance weight) that need to be adjusted simultaneously under varying operating conditions, the parameter dimensionality is insufficient and the nonlinear coupling relationships between parameters cannot be handled.

[0010] In summary, existing technologies have the following core technical defects when dealing with thermal error modeling of machine tools under varying operating conditions: (1) The machine tool is treated as a single thermal system, and the modeling is not performed according to the source and region of thermal error, making it impossible to distinguish the different contribution mechanisms of heat sources in different regions to the final error; (2) The physical model parameters in the hybrid driving method are fixed and do not have the ability to adaptively update with changes in operating conditions; (3) There is a lack of independent identification methods for thermal errors in the machining area, and the errors caused by cutting heat are confused with the thermal errors in the non-machining area; (4) There is a lack of an effective mechanism for dynamic updating of the weights of multidimensional thermal action parameters, making it difficult to balance physical interpretability and adaptive capability under varying operating conditions. Therefore, there is an urgent need for a thermal error modeling method that can realize source and region modeling of machine tool thermal errors under varying operating conditions, dynamic updating of the weights of thermal action parameters, and independent identification of thermal errors in the machining area, so as to solve the technical problems of insufficient model generalization ability, decreased prediction accuracy, or even failure of existing technologies under varying operating conditions. Summary of the Invention

[0011] In view of this, the purpose of this invention is to provide a machine tool thermal error modeling method for updating the weights of thermal parameters under varying operating conditions. By using source-regional modeling, dynamic weight updating via neural networks, and residual identification of the machining area, this method achieves high-precision dynamic modeling and prediction of machine tool thermal errors under varying operating conditions, while also taking into account the interpretability and adaptability of the model.

[0012] To achieve the above objectives, the present invention provides the following technical solution: A method for modeling machine tool thermal errors by updating the weights of thermal parameters under varying operating conditions includes: Step 1: Based on the machine tool model, divide the machine tool thermal error into sources and regions, namely, non-machining area thermal error and machining area equivalent thermal error; the non-machining area thermal error includes spindle-side thermal error, table-side thermal error and feed axis-side thermal error; divide the working conditions into internal heat source working conditions and external boundary working conditions; for any non-machining thermal error region, divide the internal heat source working conditions into primary working conditions and secondary working conditions; Step 2: Establish the thermal error in the non-processing area caused by the internal heat source conditions. Based on the characteristics of heat generation and heat transfer, the main conditions change the equivalent heat source intensity of the corresponding area, and the secondary conditions change the equivalent heat source intensity of other areas and transfer it to the target area along the heat path. Step 3: Collect temperature data of heat-sensitive points in the target area under the main operating conditions, calculate and regress to obtain the equivalent heat source intensity; Step 4: Calculate the equivalent path distance from the heat source to the target area, and calculate the thermal path coupling coefficient accordingly; Step 5: Collect temperature data of thermally sensitive points in the target area under external boundary conditions, and regress to obtain the equivalent disturbance intensity of the external boundary conditions; Step Six: Based on the thermal state time delay effect, establish a temperature response model for the target area. The temperature response model includes the regional thermal inertia coefficient, the initial weight of the equivalent heat source intensity, the initial weight of the thermal path coupling, and the initial weight of the equivalent disturbance intensity of the external boundary. It is obtained through experimental identification. The method is as follows: conduct multi-condition combination experiments, so that all conditions are combined and varied at multiple levels, collect temperature data of each non-processing thermal error area, and then identify it through least squares. Step 7: Based on the temperature response model, establish a local thermal error model for the non-processing area; Step 8: Construct a neural network for each non-processing thermal error region, and update the weights of thermal action parameters in real time based on variable operating condition information. The weights of thermal action parameters include the weights of equivalent heat source intensity, thermal path coupling, and external boundary equivalent disturbance. Step 9: Substitute the updated thermal effect parameter weights into the temperature response model to obtain the updated regional equivalent temperature response, and then output the updated local thermal error of the non-processed area. Step 10: Establish coordinate mapping matrices from each non-machining thermal error region to the tool end and workpiece end, and perform coordinate transformation based on the machine tool kinematic chain to obtain the cumulative deformation at the tool end and the cumulative deformation at the workpiece end, thereby obtaining the mapped thermal error of the non-machining region. Step 11: The difference between the measured total thermal error of cutting and the mapped thermal error of the non-machined area is used as the equivalent residual thermal error term of the machined area, and the machined area residual thermal error model is established. Step 12: Superimpose the thermal error of the non-processed area mapping with the thermal error of the processed area to obtain the total thermal error of the machine tool.

[0013] Furthermore, in step three, the method for calculating the equivalent heat source intensity is as follows: Data collection of main working conditions Below Temperature data from several heat-sensitive points were averaged and normalized to obtain sample temperature rise values. :

[0014] in: For target area Under main operating conditions The equivalent heat source intensity is as follows; For target area The Middle Heat-sensitive points in main operating conditions The temperature at which steady state is reached; For target area The Middle The initial temperature of the heat-sensitive point; For target area The Middle The maximum steady-state temperature of the heat-sensitive point; Then through the main operating conditions and temperature rise sample values The equivalent heat source intensity is obtained from the regression:

[0015] in: For target area Under main operating conditions The equivalent heat source intensity is as follows; , and To be respectively the target area Under main operating conditions The constant term coefficients, linear term coefficients, and quadratic term coefficients in the equivalent heat source intensity regression model are obtained by least squares regression fitting.

[0016] Furthermore, in step four, the heat source To the target area Equivalent path distance Represented as:

[0017] in: Indicates the first Segment structure path length; Indicates the connection type coefficient; Indicates the contact area equivalence coefficient; This represents the contact area between the two structures. This represents the maximum value within the contact area. heat source To the target area The number of path segments; The thermal path coupling coefficient Represented as:

[0018] in: This represents the total number of non-processing thermal error regions. This is a preset minimum constant.

[0019] Furthermore, in step five, the method for calculating the equivalent disturbance intensity of the external boundary condition is as follows: Collect external boundary conditions Down Temperature data from several thermally sensitive points are used to regress the equivalent disturbance intensity of the external boundary conditions.

[0020]

[0021] in: For target area Operating conditions at the external boundary The equivalent disturbance intensity below; For target area External boundary condition parameters; For target area The change in external boundary condition parameters relative to the initial values; Indicates the initial state values ​​of the external boundary condition parameters; , and Target areas Operating conditions at the external boundary The constant term coefficients, linear term coefficients, and quadratic term coefficients in the equivalent perturbation intensity regression model are obtained by least squares regression fitting. The external boundary conditions include ambient temperature and cutting fluid temperature, respectively, and the effect of ambient temperature on the target area is obtained. equivalent disturbance intensity and the effect of cutting fluid temperature on the target area equivalent disturbance intensity .

[0022] Furthermore, in step six, the temperature response model is expressed as:

[0023] in: for target area at any time Temperature response; For time step; for target area at any time The equivalent heat source intensity; heat source To the target area The thermal path coupling coefficient; for Excluding the target area at any time Other than The equivalent heat source intensity of each non-processing thermal error region ; for target area at any time At ambient temperature The equivalent disturbance intensity below; for The effect of constant cutting fluid temperature on the target area The equivalent disturbance intensity; Let be the regional thermal inertia coefficient, satisfying ; The initial weights are the equivalent heat source intensity. The initial weights for the thermal path coupling; The initial weights are the equivalent disturbance strengths of ambient temperature. The initial weights are the equivalent disturbance strengths of the cutting fluid temperature.

[0024] Furthermore, in step seven, the local thermal error model for the non-processed area is as follows:

[0025] in: , and The identification coefficient; This is the error correction factor; for Temperature response at any given time; It is the difference between the maximum and minimum temperatures at the heat-sensitive point.

[0026] Furthermore, in step eight, the equivalent heat source intensity weight... The update method is as follows:

[0027] in: The equivalent heat source intensity weight correction factor is the output of the neural network. This represents the maximum correction ratio for heat source intensity. The initial weights are the equivalent heat source intensity. The thermal path coupling weight The update method is as follows:

[0028]

[0029] in: This is a hot path coupling weight correction factor for the neural network output. The initial weights for the thermal path coupling; This represents the total number of non-processing thermal error regions. The equivalent disturbance weight of the external boundary includes the equivalent disturbance weight of ambient temperature. Equivalent disturbance to cutting fluid temperature The update method is as follows:

[0030]

[0031] in: and These are the equivalent perturbation weight correction factors for ambient temperature and cutting fluid temperature outputs of the neural network. and These are the maximum correction ratios for ambient temperature and coolant disturbance, respectively. and These are the initial weights for the equivalent disturbance intensity of ambient temperature and the equivalent disturbance intensity of cutting fluid temperature, respectively.

[0032] Furthermore, in step ten, the coordinate mapping matrix from the thermal error of each region to the tool tip is as follows:

[0033] The coordinate mapping matrix from the thermal error of each region to the end of the workpiece is as follows:

[0034] in: For the local coordinate system of the region to base coordinates rotation matrix; This represents the position of the regional error reference point in the base coordinate system. and These represent the positions of the tool and the workpiece target points in the base coordinate system, respectively. For vectors The antisymmetric matrix, used to characterize the displacement coupling effect caused by rotational motion, is expressed as:

[0035] in: , and Let them be vectors respectively Coordinate components in the X, Y, and Z directions of the base coordinate system; middle ,exist middle .

[0036] Furthermore, in step ten, the cumulative deformation of the tool tip is:

[0037] The cumulative deformation at the workpiece end is:

[0038] in: and These are the coordinate mapping matrices from the Z-axis feed axis and the spindle side to the end of the tool, respectively; , and These are the coordinate mapping matrices from the Y-axis feed axis, the X-axis feed axis, and the worktable side to the end of the workpiece, respectively. , , , and These are the local thermal error vectors for the corresponding regions; The mapping thermal error of the non-processed area is:

[0039] in: This represents the thermal error mapped to the non-processed area.

[0040] Furthermore, in step eleven, the residual thermal error model for the processing area is:

[0041] in: for Residual thermal error in the processing area at any given time; This refers to the depth of cut. This is the residual thermal inertia coefficient; , and These are the constant term coefficients, linear term coefficients, and quadratic term coefficients in the residual thermal error regression model of the processing area, respectively, obtained through least squares regression fitting.

[0042] The beneficial effects of this invention are as follows: This invention relates to a machine tool thermal error modeling method based on weighted updates of thermal parameters under varying operating conditions. The method aims to address the technical challenge of existing machine tool thermal error models failing to adaptively update thermal parameters such as heat source intensity, thermal path coupling, and external boundary disturbances under varying operating conditions. This results in insufficient model generalization ability, decreased prediction accuracy, and even model failure. By employing source- and region-based modeling, dynamic weight updates via neural networks, and residual identification in the machining area, the method achieves high-precision dynamic modeling and prediction of machine tool thermal errors under varying operating conditions, while simultaneously ensuring model interpretability and adaptability. The following technical effects have been achieved.

[0043] (1) Significantly improve the accuracy and generalization ability of thermal error prediction under variable working conditions. By modeling by source and region, thermal error is decomposed into four independent components: spindle side, worktable side, feed axis side and machining area. The thermal parameters of each region (equivalent heat source intensity weight, thermal path coupling weight, external boundary disturbance weight) can be adjusted independently according to real-time working conditions, effectively overcoming the defect of the traditional fixed parameter model in the sharp drop in accuracy under dynamic working conditions such as speed switching and feed change.

[0044] (2) Enhance the physical interpretability and engineering credibility of the model. Unlike the pure data-driven "black box" model, the physical mechanism modeling of this invention gives the thermal path coupling coefficient (based on structural path distance, connection type, and contact area) a clear engineering physical meaning; the neural network is only used to update the weights of physical parameters, rather than end-to-end mapping, and the prediction results can be traced back to the specific heat source and heat transfer link, which is convenient for engineers to understand and adjust parameters on site.

[0045] (3) The independent identification and modeling of thermal errors in the machining area were achieved for the first time. By using the residual strategy of subtracting the non-machining area mapped thermal error from the total thermal error measured in cutting, the thermal error components in the machining area caused by cutting heat, contact state and cutting fluid cooling effect were successfully separated, filling the technical gap that existing methods cannot effectively distinguish between "internal heat source error of machine tool" and "heat source error of cutting process".

[0046] (4) A complete technical closed loop of "physical modeling → dynamic update → coordinate mapping → residual compensation" is formed. By combining homogeneous coordinate transformation of the kinematic chain, the local thermal error is mapped to the tool end / workpiece end globally, so that the compensation command is accurately matched with the actual kinematic characteristics of the machine tool, and the compensation effect is better than the traditional single-point compensation method. Attached Figure Description

[0047] To make the objectives, technical solutions, and beneficial effects of this invention clearer, the following figures are provided for illustration: Figure 1 This is a flowchart of the machine tool thermal error modeling method for updating the weights of thermal parameters under varying operating conditions according to the present invention. Figure 2 This is a schematic diagram of the machine tool structure and the motion of each axis; Figure 3 A thermal path diagram showing the thermal error region from the feed shaft side heat source in the Y direction to the main shaft side. Figure 4 With spindle-side thermal error as the target area Flowchart for calculating regional thermal error.

[0048] Explanation of reference numerals in the attached figures: 10-Bed; 11-X-axis feed side; 12-X-axis slide; 13-Y-axis feed side; 14-Rotary table; 15-Worktable side; 16-Worktable; 17-Z-axis feed axis; 18-Column; 19-Z-axis slide; 20-Spindle side; 21-Spindle. Detailed Implementation

[0049] The present invention will be further described below with reference to the accompanying drawings and specific embodiments, so that those skilled in the art can better understand and implement the present invention. However, the embodiments described are not intended to limit the present invention.

[0050] This invention aims to propose a machine tool thermal error modeling method based on weighted updates of thermal parameters under varying operating conditions. First, thermal errors are categorized into 20° thermal errors on the spindle side, 15° thermal errors on the table side, thermal errors on each feed axis side, and thermal errors in the machining area. Operating conditions are further divided into internal heat source conditions and external boundary conditions. For each thermal error region, the internal heat source conditions are further decomposed into primary and secondary conditions. Then, an experimental mapping relationship between the primary internal heat source condition and the equivalent heat source intensity is established. Combining heat transfer characteristics and experiments, a thermal path coupling relationship between the heat source and the non-machining area is established. Data-driven weighting models of the external boundary conditions on the temperature and thermal error disturbances of each region are obtained. Finally, by comprehensively identifying the region's thermal inertia coefficient, equivalent heat source intensity, thermal path coupling weights, and external boundary equivalent disturbance weights, a local thermal error model for the non-machining area is established and mapped to the tool end and workpiece end through kinematic chain coordinate transformation. Simultaneously, the difference between the measured total thermal error of cutting and the mapped thermal error of the non-machining area is identified as the equivalent residual thermal error term for the machining area. Finally, the weights of thermal action parameters are updated in real time based on the information of changing operating conditions using a neural network. The weights of thermal action parameters include the weights of equivalent heat source intensity, thermal path coupling, and external boundary equivalent disturbance, so as to realize the dynamic modeling and prediction of machine tool thermal error under changing operating conditions.

[0051] Specifically, such as Figure 1 As shown in the figure, the machine tool thermal error modeling method for updating the weights of thermal action parameters under varying operating conditions in this embodiment includes the following steps.

[0052] Step 1: Based on the machine tool model, divide the machine tool thermal error into sources and areas, namely, non-machining area thermal error and machining area equivalent thermal error; the non-machining area thermal error includes spindle side 20 thermal error, table side 15 thermal error and feed axis side thermal error; divide the working conditions into internal heat source working conditions and external boundary working conditions; for any non-machining thermal error area, divide the internal heat source working conditions into primary working conditions and secondary working conditions.

[0053] like Figure 2 As shown, in this embodiment, the machine tool thermal error is divided into sources and regions based on the machine tool model, namely, spindle-side thermal error 20, table-side thermal error 15, feed axis-side thermal error, and equivalent thermal error of the machining area. The spindle-side thermal error 20, table-side thermal error 15, and feed axis-side thermal error are all non-machining area thermal errors. The operating conditions are divided into internal heat source conditions and external boundary conditions. Internal heat source conditions refer to conditions within the machine tool's motion chain that directly affect the heat generation of the heat source, such as spindle speed 21, table speed 16, and feed rate of each axis. External boundary conditions affect the machine tool through convection, such as ambient temperature and coolant temperature. Furthermore, for any thermal error region, the internal heat source conditions are further divided into primary and secondary conditions. Since the non-machining thermal error region is divided based on the positional relationship of the heat source set, it is assumed that the machine tool includes... For a given non-machining thermal error region, multiple heat sources within the same non-machining thermal error region can be equivalently represented as a single regional heat source, and this regional heat source is considered to collectively drive the thermal deformation of that directional component. For other directional components to be modeled, models are established separately using the same method; directional components that do not require modeling are set to 0. Therefore, the non-machining region of the machine tool can be equivalent to... There are three heat sources in a given region; for any target non-processing thermal error region, its corresponding heat source is used as the primary heat source, and the rest... The operating conditions corresponding to the heat source in each region can be regarded as secondary operating conditions. For example, for the thermal error region 20 on the spindle side, the spindle speed 21 directly changes the heat generated by the spindle 21. Therefore, the spindle speed 21 is the primary operating condition in this region, while the table speed 16 and the feed rate of each axis are secondary operating conditions.

[0054] Step 2: Establish the thermal error in the non-processing area caused by the internal heat source conditions. Based on the characteristics of heat generation and heat transfer, the main conditions change the equivalent heat source intensity of the corresponding area, and the secondary conditions change the equivalent heat source intensity of other areas and transfer it to the target area along the heat path.

[0055] In this embodiment, the thermal error in the non-processing area caused by the internal heat source conditions is established based on physical information weights. Specifically, the physical information includes the equivalent heat source intensity caused by the primary operating condition and the regional thermal path coupling relationship mapped by the secondary operating conditions. Taking the thermal error region on the spindle side 20 as an example, based on the characteristics of heat generation and heat transfer, the spindle speed 21 is taken as its primary operating condition, changing the equivalent heat source intensity of the spindle 21. Other internal heat source conditions are secondary operating conditions, changing the equivalent heat source intensity of other regions and transferring it to the target region along the thermal path. Therefore, this embodiment proposes to represent the influence of secondary operating conditions by a thermal path coupling coefficient determined by the structural path, connection method, and contact area.

[0056] Step 3: Collect temperature data of heat-sensitive points in the target area under the main operating conditions, and calculate and regress to obtain the equivalent heat source intensity.

[0057] In this embodiment, the method for calculating the equivalent heat source intensity is as follows: collecting data from the main operating conditions. Below Temperature data from several heat-sensitive points were averaged and normalized to obtain sample temperature rise values. :

[0058] in: For target area Under main operating conditions The equivalent heat source intensity is as follows; For target area The Middle Heat-sensitive points in main operating conditions The temperature at which steady state is reached; For target area The Middle The initial temperature of the heat-sensitive point; For target area The Middle The maximum steady-state temperature of the heat-sensitive point.

[0059] Then through the main operating conditions and temperature rise sample values The equivalent heat source intensity is obtained from the regression:

[0060] in: For target area Under main operating conditions The equivalent heat source intensity is as follows; , and To be respectively the target area Under main operating conditions The constant term coefficients, linear term coefficients, and quadratic term coefficients in the equivalent heat source intensity regression model are obtained through least squares regression fitting. The number of experiments is not less than the minimum sample size required to identify the model parameters.

[0061] In this way, each target area can be obtained. The equivalent heat source intensity includes: at the rotational speed of the main shaft 21 The equivalent heat source intensity for the 20 thermal error region on the spindle side under the main operating conditions ; with a table speed of 16 The equivalent heat source intensity of the 15 thermal error area on the workbench side under the main operating conditions ; and feed rates along the X, Y, and Z axes respectively. , and The equivalent heat source intensity of the X, Y, and Z feed shaft side thermal errors under the main operating conditions , and Step 4: Calculate the equivalent path distance from the heat source to the target area, and calculate the thermal path coupling coefficient accordingly.

[0062] In this embodiment, the heat source To the target area Equivalent path distance for:

[0063] Wherein: the Euclidean distance between any two contacting structures in space is defined as a path segment, where the heat source To the target area common Segment path, then Indicates the first Segment structure path length; This represents the connection type coefficient, which is set in stages according to the degree of connection tightness. The contact area equivalence factor is expressed as the contact area between the two structures. Obtained through normalized calculation; This represents the contact area between the two structures. This represents the maximum value within the contact area. heat source To the target area The number of path segments.

[0064] In this embodiment, the connection type coefficient The setup method is as follows: (1) Continuous solid connection, including the same casting, the same welded whole, the same solid structural component, etc. ; (2) Rigid assembly surface connections, including bolt pre-tightening connections, flange connections, motor mounting surfaces, bearing housing mounting surfaces, large-area mating surfaces, etc. ; (3) Functional motion pair connections, including guide rail pairs, lead screw-nut pairs, bearing pairs, turntable 14 support pairs, etc. ; (4) Weak contact or thermal insulation connections, including small-area supports, local gasket connections, non-metallic insulating pads, etc. .like Figure 3 As shown, the thermal error region from the Y-axis feed axis heat source to the spindle side 20 includes the first path segment L1 from the Y-axis feed axis to the column 18, the second path segment L2 from the column 18 to the spindle side 20, the third path segment L3 from the spindle side 20 to the spindle 21, the fourth path segment L4 from the spindle 21 to the Z-axis, and the fifth path segment L5 from the Z-axis to the tool.

[0065] Thus, a heat source can be obtained. To the target area thermal path coupling coefficient , is represented as:

[0066] in: This represents the total number of non-processing thermal error regions. This is a preset minimum constant. Specifically, Figure 4 middle, The heat source is located on the side of the worktable (its main operating condition is the speed of the worktable). To the target area on the main axis side The thermal path coupling coefficient; It is the heat source on the X-axis feed axis (its main operating condition is X-axis feed rate / feed rate). To the target area on the main axis side The thermal path coupling coefficient; It serves as the heat source on the Y-axis feed side (its main operating condition is the Y-axis feed speed). To the target area on the main axis side The thermal path coupling coefficient; To represent the heat source on the Z-axis feed axis (its main operating condition is the Z-axis feed speed) To the target area on the main axis side The thermal path coupling coefficient.

[0067] Step 5: Collect temperature data of thermally sensitive points in the target area under external boundary conditions, and regress to obtain the equivalent disturbance intensity of the external boundary conditions.

[0068] In this embodiment, the method for calculating the equivalent disturbance intensity of the external boundary condition is as follows: external boundary condition data is collected. Down Temperature data from several thermally sensitive points are used to regress the equivalent disturbance intensity of the external boundary conditions.

[0069]

[0070] in: For target area Operating conditions at the external boundary The equivalent disturbance intensity below; For target area External boundary condition parameters; For target area The change in external boundary condition parameters relative to the initial values; Indicates the initial state values ​​of the external boundary condition parameters; , and Target areas Operating conditions at the external boundary The constant term coefficients, linear term coefficients, and quadratic term coefficients in the equivalent disturbance intensity regression model are obtained by least squares regression fitting.

[0071] Specifically, in this embodiment, the external boundary conditions include ambient temperature and cutting fluid temperature. That is, using the above method, the effect of ambient temperature on the target area can be obtained separately. equivalent disturbance intensity and the effect of cutting fluid temperature on the target area equivalent disturbance intensity .

[0072] Step 6: Based on the thermal state time delay effect, establish a temperature response model for the target region. The temperature response model includes the regional thermal inertia coefficient, the initial weight of the equivalent heat source intensity, the initial weight of the thermal path coupling, and the initial weight of the equivalent disturbance intensity of the external boundary, which are obtained through experiments.

[0073] In this embodiment, the temperature response model is expressed as:

[0074] in: for target area at any time Temperature response; For time step; for target area at any time The equivalent heat source intensity; heat source To the target area The thermal path coupling coefficient; for Excluding the target area at any time Other than The equivalent heat source intensity of each non-processing thermal error region ; for target area at any time At ambient temperature The equivalent disturbance intensity below; for The effect of constant cutting fluid temperature on the target area The equivalent disturbance intensity; The regional thermal inertia coefficient is used to characterize the target region. The thermal response hysteresis characteristics satisfy , ; The initial weights are the equivalent heat source intensity. The initial weights for the thermal path coupling; The initial weights are the equivalent disturbance strengths of ambient temperature. This represents the initial weight for the equivalent disturbance intensity of the cutting fluid temperature. Regional thermal inertia coefficient. Initial weights of equivalent heat source intensity Initial weights for hot path coupling Initial weights of equivalent disturbance intensity at ambient temperature Initial weights of equivalent disturbance intensity to cutting fluid temperature All of these were identified through experiments. The method involved conducting multi-condition combination experiments, combining all conditions at multiple levels, collecting temperature data from each non-processing thermal error region, and then identifying them using least squares. The number of condition groups needed to be greater than 10.

[0075] Specifically, regional thermal inertia coefficient Equivalent thermal time constant of the region Between The relationship was obtained through experiments. It can then be calculated The experimental method is as follows: a step excitation is applied to the main operating conditions of the target area, and the average temperature of the target area is collected. When the temperature reaches 63.2% of the final steady-state temperature, this time is taken as the equivalent thermal time constant. .

[0076] Step 7: Based on the temperature response model, establish a local thermal error model for the non-processing area.

[0077] In this embodiment, the local thermal error model for the non-processed area is as follows:

[0078] in: , and These are the identification coefficients, obtained through least squares regression fitting; This is an error correction factor used to characterize the degree of temperature distribution non-uniformity. For the target area Contribution to thermal error correction; for Temperature response at any given time; It represents the difference between the maximum and minimum temperatures at the heat-sensitive point, characterizing the degree of temperature non-uniformity.

[0079] Following the above method, translational and angular thermal errors in different directions can be established, and can be further expressed in vector form. The thermal error models for the remaining non-processed areas are also established according to the described steps, resulting in the spindle-side thermal error model. :

[0080] Feed shaft side thermal error Represented as:

[0081] Specifically, it depends on the machine tool structure; there may be... Shaft feed shaft side thermal error , Shaft feed shaft side thermal error , Shaft feed shaft side thermal error For convenience, these are uniformly referred to as feed shaft side thermal error. .

[0082] 15 Thermal Error Models on the Workbench Side :

[0083] in: , and These are the translational thermal error components of the 20 thermal errors on the main spindle side along the X, Y, and Z axes, respectively. , and These are the angular thermal error components of the 20 thermal error region on the main spindle side around the X, Y, and Z axes, respectively. , and These are the translational thermal error components of the feed shaft side thermal error region in the X, Y, and Z directions, respectively. , and These are the angular thermal error components of the feed axis thermal error region around the X, Y, and Z axes, respectively. , and These are the translational thermal error components of the 15 thermal error regions on the worktable side in the X, Y, and Z directions, respectively. , and These are the angular thermal error components of the 15 thermal error region on the worktable side around the X, Y, and Z axes, respectively.

[0084] Step 8: Construct a neural network for each non-processing thermal error region, and update the weights of thermal action parameters in real time based on variable operating condition information. The weights of thermal action parameters include the weights of equivalent heat source intensity, thermal path coupling, and external boundary equivalent disturbance.

[0085] Specifically, the input vectors of each neural network correspond to the respective non-processing thermal error regions: [Current...] Spindle speed at time 21 ; Worktable speed 16 Feed rate of each axis , , External boundary condition variables (Specifically, this can be broken down into external boundary condition variables for each non-processing thermal error region.) , , ); Temperature response and thermal error of each non-processing thermal error region at the previous moment , .

[0086] To account for the effects of thermal inertia and historical operating conditions, the neural network input uses a length of... Within the time window, the neural network can be implemented using a multilayer perceptron, a gated recurrent unit (GRU), a long short-term memory (LSTM) network, or a one-dimensional convolutional network. Preferably, this embodiment uses a gated recurrent unit (GRU) to construct the input sequences for each region, and obtains the results through multiple output branches. , , , During training, multi-condition experimental data were used as the training set, and the regional temperature prediction error, regional thermal error prediction error, and overall machine thermal error prediction error were used as the joint loss function.

[0087] in: For the joint loss function; For regularization terms; , , , The weighting coefficients for different error terms in the loss function are manually set during training; and Indicates the target area The temperature value and thermal error value measured in the experiment; for The total thermal error measured during cutting at any given time; for The total thermal error of the machine tool predicted at any given moment; This is a discrete time point index, corresponding to physical time. ; This is an index for the non-processing thermal error region. ; and The target region obtained by updating the weights The predicted temperature response and thermal error.

[0088] Regularization term Represented as:

[0089] in: for target area at any time The equivalent heat source intensity weight; For target area Initial weights for equivalent heat source intensity; for Time of the first A heat source to the target area Hot path coupling weights; For the first A heat source to the target area Initial weights for hot path coupling; and They are respectively target area at any time The equivalent disturbance weights for ambient temperature and cutting fluid temperature; and Target areas The initial weights of the equivalent disturbance intensity of ambient temperature and the equivalent disturbance intensity of cutting fluid temperature.

[0090] Regularization term The weights used to limit the magnitude of weight corrections to the neural network output and the rate of weight change between adjacent time steps can prevent the neural network from correcting thermal parameters to values ​​that do not conform to physical meaning, while ensuring that the weights are updated smoothly as the operating conditions change.

[0091] like Figure 4 As shown, in this embodiment, the equivalent heat source intensity weight... The update method is as follows:

[0092] in: For neural network output; This represents the maximum correction ratio for heat source intensity. The initial weights are the equivalent heat source intensity.

[0093] Hot path coupling weight The update method is as follows:

[0094]

[0095] in: For neural network output; The initial weights for the thermal path coupling; This represents the total number of non-processing thermal error regions.

[0096] The equivalent disturbance weights of the external boundary include the equivalent disturbance weights of ambient temperature. Equivalent disturbance to cutting fluid temperature The update method is as follows:

[0097]

[0098] in: and For neural network output; and These are the maximum correction ratios for ambient temperature and coolant disturbance, respectively. and These are the initial weights for the equivalent disturbance intensity of ambient temperature and the equivalent disturbance intensity of cutting fluid temperature, respectively.

[0099] Step 9: Substitute the updated thermal effect parameter weights into the temperature response model to obtain the updated regional equivalent temperature response, and then output the updated local thermal error of the non-processed area.

[0100] Specifically, the updated equivalent heat source intensity weight, thermal path coupling weight, and external boundary equivalent disturbance weight are substituted into the temperature response model in step six to obtain the updated regional equivalent temperature response. Then, according to the local thermal error model of the non-processed area in step seven, the translational thermal error components in the X, Y, and Z directions and the angular thermal error components around the X, Y, and Z axes of each non-processed thermal error area are calculated, and the updated local thermal error vector of the non-processed area is output.

[0101] Step 10: Establish coordinate mapping matrices from each non-machining thermal error region to the tool end and workpiece end, and perform coordinate transformation based on the machine tool kinematic chain to obtain the cumulative deformation at the tool end and the cumulative deformation at the workpiece end, thereby obtaining the mapped thermal error of the non-machining region.

[0102] Specifically, let the machine tool base coordinates be {0}, the local coordinate system be {i}, the origin of the local coordinate system be the deformation measurement point, and the rotation matrix from {i} to {0} be... The location of the regional error reference point in the base coordinate system is: The target points for the tool and the workpiece are and Under a fixed configuration, the coordinate mapping matrix from thermal error in each region to the tool tip is:

[0103] The coordinate mapping matrix from the thermal error of each region to the end of the workpiece is as follows:

[0104] in: For the local coordinate system of the region to base coordinates rotation matrix; This represents the position of the regional error reference point in the base coordinate system. and These represent the positions of the tool and the workpiece target points in the base coordinate system, respectively. For vectors The antisymmetric matrix, used to characterize the displacement coupling effect caused by rotational motion, is expressed as:

[0105] in: , and Let them be vectors respectively Coordinate components in the X, Y, and Z directions of the base coordinate system. middle ,exist middle .

[0106] Step 11: The difference between the measured total thermal error of cutting and the mapped thermal error of the non-machined area is used as the equivalent residual thermal error term of the machined area, and the machined area residual thermal error model is established.

[0107] In this embodiment, a dual-chain coordinate mapping is established between the tool tip and workpiece tip in the non-machining area. The tool chain is bed 10 / column 18 → Z-axis slide 19 → spindle 21 → tool; the workpiece chain is bed 10 → X-axis slide 12 → Y-axis slide → worktable 16. The cumulative deformation of the tool tip is:

[0108] The cumulative deformation at the workpiece end is:

[0109] in: and These are the coordinate mapping matrices from the Z-axis feed axis 17 side and the spindle side 20 to the tool end, respectively; , and These are the coordinate mapping matrices from the Y-axis feed axis 13, the X-axis feed axis 11, and the worktable side 15 to the end of the workpiece, respectively. , , , and These are the local thermal error vectors for the corresponding regions.

[0110] The thermal error of the non-processed area mapping is:

[0111] in: This represents the thermal error mapped to the non-processed area.

[0112] Specifically, since thermal errors in the machining area are usually difficult to measure directly, the equivalent residual thermal error term, mainly caused by cutting heat in the machining area and local boundary changes, is obtained by subtracting the mapped thermal error of the non-machining area from the total measured cutting thermal error. Specifically, the total measured cutting thermal error and the mapped thermal error of the non-machining area are first calculated for a certain cutting depth:

[0113]

[0114] in: This represents the total thermal error measured during cutting. The thermal error is mapped to the non-machined area measured during cutting. and The cutting depths are respectively At that time, the cumulative deformation at the tool tip and the cumulative deformation at the workpiece tip were measured by a laser interferometer. , , , and These are the model estimates of the local thermal error vectors for the corresponding regions, calculated from the local thermal error models of each non-processed region established in step seven.

[0115] Calculate the equivalent residual:

[0116] The residual is the part that is not explained by the thermal error model of the spindle 21, table 16, and feed axis regions under different cutting depths, that is, the equivalent thermal error correction term of the machining area.

[0117] Using the depth of cut as the primary working condition in the machining area, the residual thermal error model of the machining area is regressed using residual data at all depth of cut levels:

[0118] in: for Residual thermal error in the processing area at any given time; This refers to the depth of cut. The residual thermal inertia coefficient characterizes the dynamic hysteresis characteristics of the equivalent residual thermal error in the machining area to cutting heat, external boundary disturbances, etc., and its value range is [value range missing]. ; , and These are the coefficients of the constant term, the linear term, and the quadratic term in the residual thermal error regression model of the processing area. Specifically, , , and All results were obtained through least squares regression fitting in experiments.

[0119] Of course, in other embodiments, feed rate, spindle speed 21, table speed 16, etc. can also be introduced as working condition inputs for the machining area.

[0120] Step 12: Superimpose the thermal error of the non-processed area mapping with the thermal error of the processed area to obtain the total thermal error of the machine tool, expressed as:

[0121] in: This represents the total thermal error of the machine tool.

[0122] The above-described embodiments are merely preferred embodiments provided to fully illustrate the present invention, and the scope of protection of the present invention is not limited thereto. Equivalent substitutions or modifications made by those skilled in the art based on the present invention are all within the scope of protection of the present invention. The scope of protection of the present invention is defined by the claims.

Claims

1. A method for modeling machine tool thermal errors by updating the weights of thermal parameters under varying operating conditions, characterized in that: include: Step 1: Based on the machine tool model, divide the machine tool thermal error into sources and regions, namely, non-machining area thermal error and machining area equivalent thermal error; the non-machining area thermal error includes spindle-side thermal error, table-side thermal error and feed axis-side thermal error; divide the working conditions into internal heat source working conditions and external boundary working conditions; for any non-machining thermal error region, divide the internal heat source working conditions into primary working conditions and secondary working conditions; Step 2: Establish the thermal error in the non-processing area caused by the internal heat source conditions. Based on the characteristics of heat generation and heat transfer, the main conditions change the equivalent heat source intensity of the corresponding area, and the secondary conditions change the equivalent heat source intensity of other areas and transfer it to the target area along the heat path. Step 3: Collect temperature data of heat-sensitive points in the target area under the main operating conditions, calculate and regress to obtain the equivalent heat source intensity; Step 4: Calculate the equivalent path distance from the heat source to the target area, and calculate the thermal path coupling coefficient accordingly; Step 5: Collect temperature data of thermally sensitive points in the target area under external boundary conditions, and obtain the equivalent disturbance intensity of the external boundary conditions through regression. Step Six: Based on the thermal state time delay effect, establish a temperature response model for the target area. The temperature response model includes the regional thermal inertia coefficient, the initial weight of the equivalent heat source intensity, the initial weight of the thermal path coupling, and the initial weight of the equivalent disturbance intensity of the external boundary. It is obtained through experimental identification. The method is as follows: conduct multi-condition combination experiments, so that all conditions are combined and varied at multiple levels, collect temperature data of each non-processing thermal error area, and then identify it through least squares. Step 7: Based on the temperature response model, establish a local thermal error model for the non-processing area; Step 8: Construct a neural network for each non-processing thermal error region, and update the weights of thermal action parameters in real time based on variable operating condition information. The weights of thermal action parameters include the weights of equivalent heat source intensity, thermal path coupling, and external boundary equivalent disturbance. Step 9: Substitute the updated thermal effect parameter weights into the temperature response model to obtain the updated regional equivalent temperature response, and then output the updated local thermal error of the non-processed area. Step 10: Establish coordinate mapping matrices from each non-machining thermal error region to the tool end and workpiece end, and perform coordinate transformation based on the machine tool kinematic chain to obtain the cumulative deformation at the tool end and the cumulative deformation at the workpiece end, thereby obtaining the mapped thermal error of the non-machining region. Step 11: The difference between the measured total thermal error of cutting and the mapped thermal error of the non-machined area is used as the equivalent residual thermal error term of the machined area, and the machined area residual thermal error model is established. Step 12: Superimpose the thermal error of the non-processed area mapping with the thermal error of the processed area to obtain the total thermal error of the machine tool.

2. The machine tool thermal error modeling method for updating the weights of thermal action parameters under varying operating conditions according to claim 1, characterized in that: In step three, the method for calculating the equivalent heat source intensity is as follows: Data collection of main working conditions Below Temperature data from several heat-sensitive points were averaged and normalized to obtain sample temperature rise values. : in: For target area Under main operating conditions Sample values ​​of temperature rise; For target area The Middle Heat-sensitive points in main operating conditions The temperature after reaching steady state; For target area The Middle The initial temperature of the heat-sensitive point; For target area The Middle The maximum steady-state temperature of the heat-sensitive point; Then through the main operating conditions and temperature rise sample values The equivalent heat source intensity is obtained from the regression: in: For target area Under main operating conditions The equivalent heat source intensity is as follows; , and To be respectively the target area Under main operating conditions The constant term coefficients, linear term coefficients, and quadratic term coefficients in the equivalent heat source intensity regression model are obtained by least squares regression fitting.

3. The machine tool thermal error modeling method for updating the weights of thermal action parameters under varying operating conditions according to claim 1, characterized in that: In step four, the heat source To the target area Equivalent path distance Represented as: in: Indicates the first Segment structure path length; Indicates the connection type coefficient; Indicates the contact area equivalence coefficient; This represents the contact area between the two structures. This represents the maximum value within the contact area. heat source To the target area The number of path segments; The thermal path coupling coefficient Represented as: in: This represents the total number of non-processing thermal error regions. This is a preset minimum constant.

4. The machine tool thermal error modeling method for updating the weights of thermal action parameters under varying operating conditions according to claim 1, characterized in that: In step five, the calculation method for the equivalent disturbance intensity of the external boundary condition is as follows: Collect external boundary conditions Down Temperature data from several thermally sensitive points are used to regress the equivalent disturbance intensity of the external boundary conditions. in: For target area Operating conditions at the external boundary The equivalent disturbance intensity below; For target area External boundary condition parameters; For target area The change in external boundary condition parameters relative to the initial values; Indicates the initial state values ​​of the external boundary condition parameters; , and Target areas Operating conditions at the external boundary The constant term coefficients, linear term coefficients, and quadratic term coefficients in the equivalent perturbation intensity regression model are obtained by least squares regression fitting. The external boundary conditions include ambient temperature and cutting fluid temperature, respectively, and the effect of ambient temperature on the target area is obtained. equivalent disturbance intensity and the effect of cutting fluid temperature on the target area equivalent disturbance intensity .

5. The machine tool thermal error modeling method for updating the weights of thermal action parameters under varying operating conditions according to claim 1, characterized in that: In step six, the temperature response model is expressed as: in: for target area at any time Temperature response; For time step; for target area at any time The equivalent heat source intensity; heat source To the target area The thermal path coupling coefficient; for Excluding the target area at any time Other than The equivalent heat source intensity of each non-processing thermal error region ; for target area at any time At ambient temperature The equivalent disturbance intensity below; for The effect of constant cutting fluid temperature on the target area The equivalent disturbance intensity; The regional thermal inertia coefficient satisfies ; The initial weights are the equivalent heat source intensity; The initial weights for the hot path coupling; The initial weights are the equivalent disturbance strengths of ambient temperature. The initial weights are the equivalent disturbance strengths of the cutting fluid temperature.

6. The machine tool thermal error modeling method for updating the weights of thermal action parameters under varying operating conditions according to claim 1, characterized in that: In step seven, the local thermal error model for the non-processed area is as follows: in: , and The identification coefficient; This is the error correction factor; for Temperature response at any given time; It is the difference between the maximum and minimum temperatures at the heat-sensitive point.

7. The machine tool thermal error modeling method for updating the weights of thermal action parameters under varying operating conditions according to claim 1, characterized in that: In step eight, the equivalent heat source intensity weight The update method is as follows: in: The equivalent heat source intensity weight correction factor is the output of the neural network. This represents the maximum correction ratio for heat source intensity. The initial weights are the equivalent heat source intensity; The thermal path coupling weight The update method is as follows: in: This is a hot path coupling weight correction factor for the neural network output. The initial weights for the hot path coupling; This represents the total number of non-processing thermal error regions. The equivalent disturbance weight of the external boundary includes the equivalent disturbance weight of ambient temperature. Equivalent disturbance to cutting fluid temperature The update method is as follows: in: and These are the equivalent perturbation weight correction factors for ambient temperature and cutting fluid temperature outputs of the neural network. and These are the maximum correction ratios for ambient temperature and coolant disturbance, respectively. and These are the initial weights for the equivalent disturbance intensity of ambient temperature and the equivalent disturbance intensity of cutting fluid temperature, respectively.

8. The machine tool thermal error modeling method for updating the weights of thermal action parameters under varying operating conditions according to claim 1, characterized in that: In step ten, the coordinate mapping matrix from the thermal error of each region to the tool tip is as follows: The coordinate mapping matrix from the thermal error of each region to the end of the workpiece is as follows: in: For the local coordinate system of the region to base coordinates The rotation matrix; This represents the position of the regional error reference point in the base coordinate system. and These represent the positions of the tool and the workpiece target points in the base coordinate system, respectively. For vectors The antisymmetric matrix, used to characterize the displacement coupling effect caused by rotational motion, is expressed as: in: , and Let them be vectors respectively Coordinate components in the X, Y, and Z directions of the base coordinate system; middle ,exist middle .

9. The machine tool thermal error modeling method for updating the weights of thermal action parameters under varying operating conditions according to claim 1, characterized in that: In step ten, the cumulative deformation of the tool tip is: The cumulative deformation at the workpiece end is: in: and These are the coordinate mapping matrices from the Z-axis feed axis and the spindle side to the end of the tool, respectively; , and These are the coordinate mapping matrices from the Y-axis feed axis, the X-axis feed axis, and the worktable side to the end of the workpiece, respectively. , , , and These are the local thermal error vectors for the corresponding regions; The mapping thermal error of the non-processed area is: in: This represents the thermal error mapped to the non-processed area.

10. The machine tool thermal error modeling method for updating the weights of thermal action parameters under varying operating conditions according to claim 1, characterized in that: In step eleven, the residual thermal error model for the processing area is: in: for Residual thermal error in the processing area at any given time; This refers to the depth of cut. This is the residual thermal inertia coefficient; , and These are the constant term coefficients, linear term coefficients, and quadratic term coefficients in the residual thermal error regression model of the processing area, respectively, obtained through least squares regression fitting.

Images