A method for manufacturing equipment digital twin model intelligent calibration and parameter updating

Abstract

The application discloses a kind of manufacturing equipment digital twin model intelligent calibration and parameter updating method.The method is according to the source of sensor measurement error, geometric error, thermal error, dynamic error and process error, etc., divides digital twin model into several error sub-models, and establishes calibration strategy table containing preset calibration item, allowed parameter set, parameter updating constraint, verification item, identification confidence threshold, calibration priority, etc.;According to the deviation of model prediction value and reference measurement value, the error feature vector is extracted, the error source matching degree is calculated, and the error sub-model to be calibrated and the identification confidence are obtained;According to the intelligent calibration credibility, select calibration item, and only update the parameters in the allowed parameter set;Through independent verification and verification benefit index, determine the save enablement or rollback of candidate parameters, and update the calibration priority, so as to reduce the risk of parameter misupdate when multiple error sources are coupled, and improve the long-term operation stability of the model.

Description

Technical Field

[0001] This invention belongs to the field of intelligent manufacturing technology, specifically relating to an intelligent calibration and parameter update method for digital twin models of manufacturing equipment, which is particularly applicable to the intelligent calibration and parameter update of digital twin models of manufacturing equipment such as ultra-precision machine tools and CNC machining equipment. Background Technology

[0002] Digital twin models of manufacturing equipment can predict the state of manufacturing equipment, the state of the manufactured object, machining errors, thermal drift, dynamic response, or process results based on equipment operating variables, environmental variables, motion variables, or process variables. For manufacturing equipment such as ultra-precision machine tools, CNC machining equipment, grinding equipment, and polishing equipment, digital twin models can provide a model basis for error prediction, state assessment, process optimization, and compensation control.

[0003] However, during long-term operation, the actual state of manufacturing equipment is affected by factors such as sensor drift, changes in installation status, changes in geometric relationships, changes in temperature fields, changes in structural dynamic characteristics, changes in process loads, wear of cutting tools or grinding wheels, and changes in the state of the manufactured object. This causes the predicted values ​​of the digital twin model to gradually deviate from the actual measured values. If the digital twin model is not calibrated and its parameters are not updated in a timely manner, the model prediction error will continue to accumulate, thereby affecting the accuracy of equipment status judgment, error compensation, and process control.

[0004] Existing digital twin model calibration methods typically employ offline calibration, periodic model reconstruction, or direct adjustment of model parameters based on prediction errors. While these methods can improve model accuracy to some extent, they still suffer from the following problems: First, existing methods are insufficient in distinguishing the sources of model prediction errors. Prediction errors in manufacturing equipment digital twin models may originate from different aspects such as sensor measurements, geometric mapping, thermal deformation, dynamic response, or technological processes. If model parameters are directly adjusted based solely on the overall deviation between the model's predicted values ​​and reference measurements, errors from different sources are easily mixed, leading to inaccurate judgments of the object to be calibrated. Second, existing methods lack sufficient constraints on the objects to be updated with parameters. Different error sources correspond to different calibration items, reference measurement objects, and updatable parameters. Updating model parameters without clearly defining the error source, calibration items, and the set of allowed update parameters can easily result in irrelevant parameters being mistakenly updated, causing parameter drift or localized accuracy degradation in the digital twin model. Third, existing methods lack sufficient constraints on the reliability of calibration triggering conditions. When the excitation range is insufficient, the repeatability of reference measurements is poor, or the error variation pattern is inconsistent with the target error source, if parameter updates are still performed, it is easy to mistakenly identify deviations caused by accidental measurement fluctuations or non-target error sources as model parameter mismatch. Fourth, existing methods lack sufficient verification and feedback of update results. Some methods directly use the original calibration data to evaluate the update effect after parameter updates, lacking independent verification, rollback, and subsequent calibration priority adjustment mechanisms. When candidate parameters perform poorly under other measurement locations, operating states, temperature ranges, load levels, or process parameters, it is difficult to restore the model parameters before the update in a timely manner, affecting the long-term stability of the digital twin model.

[0005] Therefore, there is an urgent need to propose an intelligent calibration and parameter update method for digital twin models of manufacturing equipment. This method can improve the pertinence, stability and reliability of parameter calibration of digital twin models by extracting error features, matching and identifying error sources, matching calibration strategies, judging calibration credibility, updating restricted parameters within the allowed parameter set, and establishing an independent verification rollback mechanism. In particular, it can reduce the risk of incorrect parameter updates under multiple error source coupling conditions. Summary of the Invention

[0006] This invention addresses the problems of difficulty in accurately distinguishing error sources, easy erroneous or complete updating of model parameters, insufficient reliability of calibration triggering conditions, and lack of reliable verification and feedback of update results in the calibration process of existing digital twin models of manufacturing equipment. It provides an intelligent calibration and parameter update method for digital twin models of manufacturing equipment.

[0007] To solve the above-mentioned technical problems, the present invention adopts the following technical solution:

[0008] This invention provides a method for intelligent calibration and parameter updating of a digital twin model of manufacturing equipment, the method comprising:

[0009] Step S1: Establish a digital twin model of the manufacturing equipment. The digital twin model of the manufacturing equipment is used to output model prediction values ​​corresponding to the state of the manufacturing equipment or the state of the manufactured object based on the operating variables, environmental variables, motion variables or process variables of the manufacturing equipment.

[0010] Step S2: Based on the impact of different error sources on the model's predicted values, the digital twin model of the manufacturing equipment is divided into several error sub-models;

[0011] Step S3: Establish a calibration strategy table, which is used to record the correspondence between error sub-models, preset calibration items, allowed parameter update sets, parameter update constraints, verification items, historical error feature templates, feature weights, identification confidence thresholds, intelligent calibration confidence thresholds, and calibration priorities.

[0012] Step S4: Obtain the deviation between the model prediction value and the reference measurement value of the digital twin model of the manufacturing equipment;

[0013] Step S5: Extract the error feature vector based on the deviation and the repeatability of the corresponding operating variables, environmental variables, motion variables, process variables, or reference measurement values;

[0014] Step S6: Based on the error feature vector and the historical error feature template and feature weights corresponding to each error sub-model, calculate the error source matching degree of each error sub-model to obtain the error sub-model to be calibrated and the recognition confidence degree corresponding to the error sub-model to be calibrated.

[0015] Step S7: When the identification confidence level reaches the corresponding identification confidence level threshold in the calibration strategy table, select the corresponding preset calibration item according to the error sub-model to be calibrated, identification confidence level, error source matching degree and calibration strategy table, and control the manufacturing equipment to execute the preset calibration item; when the identification confidence level does not reach the corresponding identification confidence level threshold, keep the parameters of the manufacturing equipment digital twin model unchanged, and output a supplementary calibration or manual review prompt;

[0016] Step S8: During the process of the manufacturing equipment performing the preset calibration items, the actual values ​​of the controlled variable or excitation variable and the reference measurement value corresponding to the reference measurement object are obtained, and the calibration input variables, including the actual values ​​of the controlled variable or excitation variable, are input into the digital twin model of the manufacturing equipment to obtain the model prediction value corresponding to the reference measurement value, and the calibration error between the model prediction value and the reference measurement value is calculated.

[0017] Step S9: Calculate the intelligent calibration confidence based on the identification confidence, error source matching degree, target error sub-model corresponding to the preset calibration item, excitation coverage of the controlled variable or excitation variable, repeatability of the reference measurement value, error pattern matching factor and calibration error, and determine whether the preset calibration item meets the intelligent calibration conditions based on the intelligent calibration confidence.

[0018] Step S10: When the preset calibration item meets the intelligent calibration conditions, calculate the candidate update amount of the digital twin model parameters within the allowed update parameter set, parameter sensitivity matrix, and parameter update constraints in the preset calibration item, and keep the digital twin model parameters that do not belong to the allowed update parameter set unchanged to obtain the candidate digital twin model parameters; when the preset calibration item does not meet the intelligent calibration conditions, keep the digital twin model parameters of the manufacturing equipment unchanged.

[0019] Step S11: After obtaining the candidate digital twin model parameters, execute the verification project, obtain the reference measurement value corresponding to the verification project, and calculate the updated verification error and the original verification error based on the candidate digital twin model parameters and the original digital twin model parameters, respectively. Then, determine the verification benefit index based on the updated verification error, the original verification error, and the candidate update amount.

[0020] Step S12: Determine the parameter processing method based on the verification benefit index, the updated verification error, the verification error before the update, the parameter update constraints, and the historical verification results. Save and enable the candidate digital twin model parameters or roll back to the digital twin model parameters before the update according to the parameter processing method, and update the calibration priority of the preset calibration items.

[0021] Furthermore, the error sub-model in step S2 includes two or more of the following: sensor measurement error sub-model, geometric error sub-model, thermal error sub-model, dynamic error sub-model, and process error sub-model.

[0022] The sensor measurement error sub-model is used to characterize the impact of sensor zero-point offset, sensitivity variation, range deviation, or measurement noise on the model's predicted values.

[0023] The geometric error sub-model is used to characterize the influence of coordinate mapping error, installation pose error or motion geometric error between the equipment coordinate system, sensor coordinate system, motion actuator coordinate system and manufacturing object coordinate system on the model prediction value.

[0024] The thermal error sub-model is used to characterize the effects of temperature changes, heat source power changes, thermal hysteresis, or thermal deformation on the model's predicted values.

[0025] The dynamic error sub-model is used to characterize the impact of changes in rotational speed, velocity, acceleration, vibration, damping, or dynamic stiffness on the model's predicted values.

[0026] The process error sub-model is used to characterize the influence of force, power, current, contact state, or material removal on the model's predicted values ​​during cutting, grinding, polishing, or surface treatment.

[0027] Furthermore, the calibration strategy table in step S3 includes several preset calibration items. Each preset calibration item corresponds to the target error sub-model, calibration execution method, controlled variable or excitation variable, reference measurement object, set of parameters that can be updated, parameter update constraints, verification items, historical error feature template, feature weight, identification confidence threshold, intelligent calibration confidence threshold, and calibration priority.

[0028] The target error sub-model is used to represent the error source corresponding to the preset calibration item;

[0029] The calibration execution mode is used to indicate the motion mode, operation mode, or process execution mode of the manufacturing equipment when performing the preset calibration item;

[0030] The controlled variable or excitation variable is used to represent the variable that is actively changed or controlled to be maintained when the preset calibration item is performed;

[0031] The reference measurement object is used to compare with the model prediction value output by the digital twin model of the manufacturing equipment;

[0032] The set of allowed update parameters is used to limit the digital twin model parameters that can be updated when the preset calibration item meets the intelligent calibration conditions;

[0033] The parameter update constraint is used to limit the maximum single update amount, minimum update interval, upper limit of cumulative change, or range of parameter values ​​within the parameter set that are allowed to be updated.

[0034] The historical error feature template and feature weights are used to calculate the error source matching degree between the current error feature vector and each error sub-model;

[0035] The verification project is used to verify the parameters of the candidate digital twin model after the parameters are updated;

[0036] The calibration priority is determined based on the importance of the error sub-model corresponding to the preset calibration item, the number of successful historical calibrations, the number of historical rollbacks, the time of the most recent calibration, or the verification benefit results.

[0037] Furthermore, in step S5, the error feature vector includes two or more of the following features:

[0038] Error magnitude characteristics are used to characterize the magnitude of the deviation between the model's predicted values ​​and the reference measured values;

[0039] Repeatability characteristics are used to characterize the degree of dispersion of reference measurement values ​​at the same measurement location, under the same standard input value, or under the same operating conditions.

[0040] Position-related features are used to characterize the degree to which deviation changes with measurement position, direction of motion, trajectory point number, or reciprocating motion direction;

[0041] Temperature-dependent characteristics are used to characterize the degree to which deviations vary with temperature, heat source power, operating time, or cooling time.

[0042] Dynamic correlation characteristics are used to characterize the degree to which deviation changes with rotational speed, velocity, acceleration, excitation frequency, or motion reversal state;

[0043] Process-related characteristics are used to characterize the extent to which deviations vary with process parameters, load levels, power, current, or contact conditions.

[0044] Furthermore, the historical error feature template is determined based on the historical error feature vector in the historical calibration samples, and the feature weights are determined based on the statistical results of the historical calibration samples or the physical mechanisms corresponding to each error sub-model.

[0045] Further, in step S6, obtaining the error sub-model to be calibrated and identifying the confidence level includes:

[0046] Step S61: Obtain historical calibration samples, which include historical error feature vectors and corresponding error source labels. The error source labels include one of the following: sensor measurement error label, geometric error label, thermal error label, dynamic error label, and process error label.

[0047] Step S62: Based on the historical error feature vectors belonging to the same error source label, form the historical error feature template of the corresponding error sub-model, and determine the feature weights corresponding to the error sub-model;

[0048] Step S63: Based on the current error feature vector, historical error feature templates, and feature weights, calculate the error source matching degree between the current error feature vector and each error sub-model;

[0049] Step S64: The error sub-model with the highest error source matching degree is determined as the error sub-model to be calibrated, and the error source matching degree or the value obtained based on the error source matching degree is used as the identification confidence level;

[0050] Step S65: When the recognition confidence level is lower than the corresponding recognition confidence level threshold in the calibration strategy table, keep the parameters of the manufacturing equipment digital twin model unchanged and output a supplementary calibration or manual review prompt.

[0051] Further, in step S7, selecting the corresponding preset calibration items based on the error sub-model to be calibrated, the identification confidence level, the error source matching degree, and the calibration strategy table includes:

[0052] Step S71: In the calibration strategy table, select preset calibration items whose target error sub-model is consistent with the error sub-model to be calibrated;

[0053] Step S72: From the selected preset calibration items, retain the preset calibration items that have reached the corresponding recognition confidence threshold and have controlled variables or excitation variables and reference measurement objects that match the current operating status of the manufacturing equipment.

[0054] Step S73: Select the preset calibration item with the highest calibration priority from the reserved preset calibration items according to the calibration priority, error source matching degree, number of historical successful verifications, number of historical rollbacks, or the time of the most recent calibration item.

[0055] Step S74: When there is no preset calibration item whose recognition confidence reaches the corresponding recognition confidence threshold, or when there is no preset calibration item that matches the current operating status of the manufacturing equipment, keep the parameters of the digital twin model of the manufacturing equipment unchanged, and output a prompt for supplementary calibration or manual review.

[0056] Furthermore, in step S8, the preset calibration items, controlled variables or excitation variables, and reference measurement values ​​corresponding to different error sub-models include:

[0057] When the preset calibration items correspond to the sensor measurement error sub-model, the preset calibration items include sensor zero-position measurement items, repeated measurement items at the same position on a fixed reference surface, or standard value input measurement items; the controlled variable or excitation variable includes standard input value, number of repeated measurements, or sensor zero-position state; the reference measurement value includes sensor zero-position measurement value, repeated measurement value on a fixed reference surface, or sensor output value under standard value input conditions;

[0058] When the preset calibration items correspond to the geometric error sub-model, the preset calibration items include multi-position reference measurement items, standard trajectory scanning items, or bidirectional reciprocating positioning measurement items; the controlled variables or excitation variables include the position of the motion actuator, the direction of motion, the trajectory point number, or the direction of reciprocating motion; the reference measurement values ​​include reference component measurement values, standard trajectory measurement values, or multi-position geometric measurement values;

[0059] When the preset calibration items correspond to the thermal error sub-model, the preset calibration items include no-load heating measurement items, constant-rate thermal stability measurement items, or cooling process measurement items; the controlled variable or excitation variable includes temperature, heat source power, running time, or cooling time; the reference measurement value includes thermal drift measurement value or displacement measurement value under temperature change conditions;

[0060] When the preset calibration items correspond to the dynamic error sub-model, the preset calibration items include variable speed motion measurement items, step motion measurement items, reciprocating acceleration and deceleration measurement items, or dynamic excitation measurement items; the controlled variable or excitation variable includes rotational speed, velocity, acceleration, excitation frequency, or motion reversal state; the reference measurement values ​​include vibration response measurement values, dynamic displacement measurement values, or motion response measurement values.

[0061] When the preset calibration items correspond to the process error sub-model, the preset calibration items include process parameter operation measurement items, load level operation measurement items, or energy input operation measurement items; the controlled variables or excitation variables include process parameters, load levels, power, current, or contact status; the reference measurement values ​​include process load measurement values, post-processed manufactured object measurement values, or material removal measurement values.

[0062] Further, in step S9, determining whether the preset calibration item meets the intelligent calibration conditions includes:

[0063] Step S91: Determine whether the target error sub-model corresponding to the preset calibration item is consistent with the error sub-model to be calibrated;

[0064] Step S92: Determine whether the recognition confidence level reaches the recognition confidence threshold corresponding to the preset calibration item;

[0065] Step S93: Determine whether the actual value of the controlled variable or the excitation variable reaches the preset range of change corresponding to the target error sub-model, and determine the excitation coverage based on the relationship between the actual value and the preset range of change;

[0066] Step S94: Determine whether the repeatability of the reference measurement value meets the preset repeatability requirement, and determine the reference measurement repeatability factor based on the dispersion of the reference measurement value;

[0067] Step S95: Determine whether the calibration error exceeds a preset error threshold, and determine the error amplitude validity factor based on the relationship between the calibration error and the preset error threshold;

[0068] Step S96: Determine whether the characteristics of the calibration error changing with the controlled variable or the excitation variable match the error change characteristics corresponding to the target error sub-model, and determine the error mode matching factor based on the degree of matching of the error source or the degree of consistency of the error change characteristics;

[0069] Step S97: Determine the confidence level of intelligent calibration based on the identification confidence level, excitation coverage, reference measurement repeatability factor, error pattern matching factor, and error amplitude validity factor;

[0070] When the target error sub-model corresponding to the preset calibration item is consistent with the error sub-model to be calibrated, and the smart calibration confidence level reaches the corresponding smart calibration confidence level threshold, the preset calibration item is determined to meet the smart calibration conditions; when the target error sub-model corresponding to the preset calibration item is inconsistent with the error sub-model to be calibrated, or the smart calibration confidence level does not reach the corresponding smart calibration confidence level threshold, the preset calibration item is determined to not meet the smart calibration conditions.

[0071] Furthermore, in step S10, updating only the digital twin model parameters within the set of allowed update parameters in the preset calibration project includes:

[0072] Step S101: Based on the target error sub-model corresponding to the preset calibration item, determine the set of parameters that can be updated in the calibration strategy table;

[0073] Step S102: Based on the actual values ​​of the controlled or excitation variables collected during the calibration process, calculate the local sensitivity of the model prediction values ​​to each parameter within the allowed update parameter set, and obtain the parameter sensitivity matrix;

[0074] Step S103: Based on the calibration error and parameter sensitivity matrix, calculate the candidate update amounts for the parameters of the digital twin model within the allowed parameter set;

[0075] Step S104: Determine whether the candidate update amount meets the requirements of maximum single update amount, minimum update interval, upper limit of cumulative change amount, and parameter value range based on the parameter update constraints;

[0076] Step S105: When the candidate update quantity satisfies the parameter update constraint, candidate digital twin model parameters are generated; when the candidate update quantity does not satisfy the parameter update constraint, the candidate update quantity is not enabled, and the digital twin model parameters before the update remain unchanged.

[0077] When the preset calibration item corresponds to the sensor measurement error sub-model, the set of parameters that can be updated includes sensor zero-point parameters, sensitivity parameters, range correction parameters, or measurement noise parameters.

[0078] When the preset calibration item corresponds to the geometric error sub-model, the set of parameters that can be updated includes coordinate mapping parameters, installation posture parameters, measurement direction parameters, motion axis positioning error parameters, straightness error parameters, or backlash parameters.

[0079] When the preset calibration item corresponds to the thermal error sub-model, the set of allowed update parameters includes the temperature sensitivity coefficient, thermal hysteresis time, thermal drift mapping coefficient, or thermal boundary correction coefficient.

[0080] When the preset calibration item corresponds to the dynamic error sub-model, the set of parameters that can be updated includes dynamic stiffness parameters, damping parameters, vibration transmission parameters, rotational speed related response parameters, or acceleration / deceleration response parameters.

[0081] When the preset calibration item corresponds to the process error sub-model, the set of allowed update parameters includes process load mapping parameters, contact stiffness parameters, power mapping parameters, current mapping parameters, or material removal effect parameters.

[0082] For digital twin model parameters that do not belong to the set of allowed update parameters, their candidate update values ​​are not calculated and the parameter values ​​are kept unchanged.

[0083] Further, steps S11 and S12 include:

[0084] Step S111: After obtaining the parameters of the candidate digital twin model, select a verification item. The verification item is different from the preset calibration item used for parameter update; or, the verification item and the preset calibration item used for parameter update belong to the same type of preset calibration item, but their measurement positions, motion trajectories, running speeds, temperature ranges, load levels, process parameters, or running time periods are different.

[0085] Step S112: Execute the verification item and obtain the reference measurement value corresponding to the verification item;

[0086] Step S113: Calculate the updated verification error and the original verification error based on the candidate digital twin model parameters and the original digital twin model parameters, respectively;

[0087] Step S114: Determine the verification benefit index corresponding to the candidate digital twin model parameters based on the updated verification error, the original verification error, and the candidate update amount within the allowed update parameter set;

[0088] Step S121: Determine whether the verification benefit index has reached the preset verification benefit threshold;

[0089] Step S122: Based on the parameter update constraints, determine whether the parameters of the candidate digital twin model meet the requirements of maximum single update amount, minimum update interval, upper limit of cumulative change amount, or parameter value range.

[0090] Step S123: Based on the historical verification results, determine whether the same model parameters, the same error sub-model, or the same preset calibration item have reached the preset historical rollback limit;

[0091] Step S124: When the verification benefit index reaches the preset verification benefit threshold, the candidate digital twin model parameters meet the parameter update constraints, and the preset historical rollback limit is not reached, the parameter processing method is determined to be save and enable, the candidate digital twin model parameters are saved and enabled, and a new digital twin model version is generated.

[0092] Step S125: When the verification benefit index does not reach the preset verification benefit threshold, or the candidate digital twin model parameters do not meet the parameter update constraints, or the preset historical rollback limit is reached, the parameter processing method is determined to be rollback, rollback to the digital twin model parameters before the update, and the rollback reason is recorded.

[0093] Step S126: Update the calibration priority of the preset calibration item according to the parameter processing method. When the parameter processing method is "save enabled", increase the calibration priority of the corresponding preset calibration item or record a historical calibration success count. When the parameter processing method is "rollback", decrease the calibration priority of the corresponding preset calibration item or record a historical rollback count.

[0094] Step S127: When the number of historical rollbacks for the same model parameter, the same error sub-model, or the same preset calibration item reaches the preset number, pause the automatic parameter update of the corresponding model parameter, error sub-model, or preset calibration item, and output a prompt for recalibration or manual review.

[0095] Compared with the prior art, the present invention has the following advantages and significant effects:

[0096] (1) To address the problem of insufficient differentiation of model prediction error sources in existing technologies, this paper extracts error feature vectors based on the deviation between model predictions and reference measurements. Then, it obtains the error sub-model to be calibrated and its identification confidence level through an error source identification model, enabling digital twin model calibration to be based on the error source identification results. By using error source matching degree, the current error feature vector is weighted and matched with the historical error feature templates of each error sub-model, enhancing the correspondence between the error source identification process and the error mechanism of manufacturing equipment. Therefore, it avoids directly updating digital twin model parameters without distinguishing between sensor measurement errors, geometric errors, thermal errors, dynamic errors, or process errors, thereby reducing misjudgments of the object to be calibrated and improving the accuracy of the judgment.

[0097] (2) To address the insufficient limitation of parameter update targets in existing technologies, a calibration strategy table is established, corresponding to the error sub-model, preset calibration items, allowed update parameter set, parameter update constraints, and identification confidence threshold. The digital twin model parameters within the allowed update parameter set are updated only when the intelligent calibration conditions are met. By comprehensively constraining the confidence level of intelligent calibration, identifying confidence, excitation coverage, reference measurement repeatability, error pattern matching factor, and error amplitude validity, parameter updates can be avoided when the excitation range is insufficient, measurement data is unstable, or the error variation pattern does not conform to the target error sub-model. Therefore, it is possible to avoid incorrectly assigning calibration errors caused by a certain error source to irrelevant model parameters, reducing the risk of parameter drift and erroneous updates.

[0098] (3) To address the problem of existing technologies lacking error source constraints in parameter update calculations, this invention calculates candidate update quantities based on the parameter sensitivity matrix within the allowed update parameter set, and restricts the candidate update quantities through parameter update constraints such as the maximum single update quantity, parameter value range, and upper limit of cumulative change. For digital twin model parameters that do not belong to the allowed update parameter set, their parameter values ​​remain unchanged. This avoids problems such as mistakenly updating geometric mapping parameters due to thermal drift errors, thus improving the targeting of digital twin model parameter updates.

[0099] (4) To address the shortcomings of existing technologies in verifying and providing feedback on updated results, a verification project is executed after generating candidate digital twin model parameters. The verification error after the update and the verification error before the update are calculated respectively. Based on the verification results, parameter update constraints, and historical verification results, the parameter processing method of saving and enabling or rolling back is determined, and the calibration priority of the corresponding preset calibration project is adjusted. By examining both the degree of reduction in verification error and the magnitude of parameter update through the verification benefit index, the problem of candidate parameters only reducing errors on the original calibration data but declining predictive performance under other measurement locations, operating states, temperature ranges, or process parameters can be avoided, thereby improving the long-term stability and reliability of the digital twin model. Attached Figure Description

[0100] Appendix Figure 1 This is a flowchart of a method for intelligent calibration and parameter updating of digital twin models of manufacturing equipment.

[0101] Appendix Figure 2 This is a schematic diagram of the error sub-model division of the digital twin model of manufacturing equipment.

[0102] Appendix Figure 3 This is a schematic diagram of the calibration strategy table.

[0103] Appendix Figure 4 This is a schematic diagram of the error feature vector extraction and error source identification model.

[0104] Appendix Figure 5 This is a flowchart of the preset calibration item selection and intelligent calibration condition judgment process.

[0105] Appendix Figure 6 It is a flowchart of the candidate digital twin model parameter verification, saving and enabling, rollback and calibration priority feedback process. Detailed Implementation

[0106] To facilitate understanding by those skilled in the art, the following description is provided in conjunction with the appendix. Figure 1-6 The present invention will be described in detail below.

[0107] As attached Figure 1 As shown, this invention provides a method for intelligent calibration and parameter updating of a digital twin model of manufacturing equipment. The manufacturing equipment can be an ultra-precision machine tool, CNC machining equipment, grinding equipment, polishing equipment, or other manufacturing equipment with motion actuators, process execution components, and state measurement units.

[0108] S1: Establish a digital twin model of manufacturing equipment. The digital twin model of manufacturing equipment is used to output model prediction values ​​corresponding to the state of manufacturing equipment or the state of manufacturing object based on the operating variables, environmental variables, motion variables or process variables of the manufacturing equipment.

[0109] In this embodiment, the operating variables may include spindle speed, feed rate, running time, equipment status variables, etc.; the environmental variables may include ambient temperature, internal equipment temperature, heat source power, etc.; the motion variables may include the position of the motion actuator, direction of motion, trajectory point number, speed, and acceleration, etc.; the process variables may include process parameters, load level, power, current, contact state, etc. The model predicted values ​​may include displacement, temperature, vibration response, machining load, material removal amount, manufacturing object morphology error, or other predicted results related to the state of the manufacturing equipment and the state of the manufacturing object.

[0110] S2: Based on the impact of different error sources on the model's predicted values, the digital twin model of the manufacturing equipment is divided into several error sub-models;

[0111] As attached Figure 2 As shown in this embodiment, the digital twin model of the manufacturing equipment is divided into several error sub-models based on the impact of different error sources on the model's predicted values. The error sub-models include two or more of the following: sensor measurement error sub-model, geometric error sub-model, thermal error sub-model, dynamic error sub-model, and process error sub-model.

[0112] The sensor measurement error sub-model characterizes the impact of sensor zero-point offset, sensitivity variation, range deviation, or measurement noise on the model's predicted values. The geometric error sub-model characterizes the impact of coordinate mapping errors, installation pose errors, or motion geometric errors between the equipment coordinate system, sensor coordinate system, motion actuator coordinate system, and manufactured object coordinate system on the model's predicted values. The thermal error sub-model characterizes the impact of temperature changes, heat source power changes, thermal hysteresis, or thermal deformation on the model's predicted values. The dynamic error sub-model characterizes the impact of changes in rotational speed, velocity, acceleration, vibration, damping, or dynamic stiffness on the model's predicted values. The process error sub-model characterizes the impact of forces, power, current, contact states, or material removal during cutting, grinding, polishing, or surface treatment processes on the model's predicted values.

[0113] Through the above division, the prediction error of the digital twin model is no longer regarded as a single, overall error, but can be identified, calibrated, and updated according to the error source. Especially when sensor measurement errors, geometric errors, thermal errors, dynamic errors, and process errors coexist or are coupled with each other, the above error sub-model division can provide an object basis for subsequent error source matching, calibration item selection, and restricted updates within the allowable parameter set.

[0114] S3: Establish a calibration strategy table, which is used to record the correspondence between error sub-model, preset calibration items, allowed update parameter set, parameter update constraints, verification items, historical error feature template, feature weight, identification confidence threshold, intelligent calibration confidence threshold and calibration priority;

[0115] As attached Figure 3 As shown in this embodiment, the calibration strategy table includes several preset calibration items. Each preset calibration item corresponds to a target error sub-model, calibration execution method, controlled variable or excitation variable, reference measurement object, set of parameters that can be updated, parameter update constraints, verification items, historical error feature template, feature weight, identification confidence threshold, intelligent calibration credibility threshold, and calibration priority.

[0116] The target error sub-model represents the error source corresponding to the preset calibration item; the calibration execution method represents the motion, operation, or process execution method of the manufacturing equipment when performing the preset calibration item; the controlled variable or excitation variable represents the variable that is actively changed or controlled and maintained when performing the preset calibration item; the reference measurement object is used to compare with the model prediction value output by the digital twin model of the manufacturing equipment; the allowed update parameter set limits the digital twin model parameters that can be updated when the preset calibration item meets the intelligent calibration conditions; the parameter update constraint limits the maximum single update amount, minimum update interval, upper limit of cumulative change, or parameter value range of the parameters in the allowed update parameter set; and the verification item is used to verify the parameters after the parameter update. Candidate digital twin model parameters are validated; historical error feature templates are used to characterize the typical error change patterns of the corresponding error sub-model in historical calibration samples; feature weights are used to characterize the importance of error amplitude features, repeatability features, location-related features, temperature-related features, dynamic-related features, and process-related features in the identification of the corresponding error sub-model; identification confidence thresholds are used to determine whether the error source identification results are credible; intelligent calibration confidence thresholds are used to determine whether the preset calibration items meet the intelligent calibration conditions; calibration priority is used for calibration item screening and order adjustment, wherein the calibration priority is determined based on the importance of the error sub-model corresponding to the preset calibration item, the number of historical calibration successes, the number of historical rollbacks, the time of the most recent calibration, or the validation benefit results.

[0117] The calibration priority can be represented by a numerical value, with a larger value indicating a higher calibration priority.

[0118] The calibration strategy table can be stored in the form of a data table, configuration file, database table, or model parameter configuration table. The identification confidence threshold and calibration priority in the calibration strategy table can be preset or updated based on subsequent validation results.

[0119] Calibration priority is used to prioritize preset calibration items that have a high historical validation success rate, a low number of historical rollbacks, high validation benefits, or a significant impact on the prediction error of the current model when multiple preset calibration items meet the screening criteria.

[0120] S4: Obtain the deviation between the model prediction value and the reference measurement value of the digital twin model of the manufacturing equipment;

[0121] As attached Figure 4 As shown, in this embodiment, an error feature vector is extracted based on the deviation and the repeatability of the corresponding operating variables, environmental variables, motion variables, process variables, or reference measurement values.

[0122] The deviation of the i-th sampling point can be expressed as:

[0123] (1)

[0124] in, This represents the deviation at the i-th sampling point. This represents the reference measurement value of the i-th sampling point. This represents the model prediction value for the i-th sampling point.

[0125] The calibration error can be obtained based on the deviation of multiple sampling points. This calibration error can be expressed as the root mean square error.

[0126] (2)

[0127] in, The root mean square error (RMSE) represents the calibration error, and N represents the number of sampling points. Besides the RMS error, other indicators that characterize the deviation between the model's predicted values ​​and the reference measurements can also be used, such as mean absolute error, maximum absolute error, weighted error, or others.

[0128] S5: Extract the error feature vector based on the deviation and the repeatability of the corresponding operating variables, environmental variables, motion variables, process variables or reference measurement values;

[0129] In this embodiment, the error feature vector can be represented as:

[0130] (3)

[0131] in, Represents the error eigenvector. Indicates the characteristics of error magnitude. Indicates repeatability. Indicates location-related features, Indicates temperature-related characteristics. Indicates dynamic correlation characteristics, Indicates process-related characteristics.

[0132] Error amplitude characteristics Used to characterize the magnitude of the deviation between the model's predicted value and the reference measurement value. It can be obtained from the root mean square error, that is:

[0133] (4)

[0134] Repeatability It is used to characterize the degree of dispersion of reference measurement values ​​at the same measurement location, under the same standard input value, or under the same operating conditions. It can be represented as:

[0135] (5)

[0136] Where M represents the number of repeated measurements, This represents the reference measurement value for the j-th repeated measurement. This represents the average of M reference measurements.

[0137] Location-related features Temperature-related characteristics Dynamic correlation characteristics Process-related characteristics These are used to characterize the degree of correlation between deviations and location-related variables, thermally related variables, dynamically related variables, and process-related variables, respectively. These characteristics can be expressed using the absolute value of the correlation coefficient:

[0138] (6)

[0139] (7)

[0140] (8)

[0141] (9)

[0142] in, represents the correlation coefficient; e represents the deviation sequence; x represents positional variables such as measurement location, direction of motion, or trajectory point number; T represents thermally related variables such as temperature, heat source power, running time, or cooling time; v represents dynamically related variables such as rotational speed, speed, acceleration, excitation frequency, or motion reversal state; q represents process-related variables such as process parameters, load level, power, current, or contact state.

[0143] The above method of constructing the error feature vector is only one option. In practical applications, the error feature vector can also be constructed using the slope of change, piecewise change, fitting residual, frequency domain features, time domain statistical features, or other features that can reflect the law of error change.

[0144] S6: Based on the error feature vector and the historical error feature template and feature weights corresponding to each error sub-model, calculate the error source matching degree of each error sub-model to obtain the error sub-model to be calibrated and the recognition confidence degree corresponding to the error sub-model to be calibrated.

[0145] In this embodiment, S6 includes the following steps:

[0146] S61: Obtain historical calibration samples, which include historical error feature vectors and corresponding error source labels. The error source labels include one of the following: sensor measurement error label, geometric error label, thermal error label, dynamic error label, and process error label.

[0147] In this embodiment, the error source identification model can be established using historical calibration samples. The error source identification model can be established using a classification model, statistical discriminant model, rule model, regression model, or machine learning model. The specific model form is not limited; any model that can output the matching degree, probability value, similarity, or other numerical values ​​that can represent the reliability of the identification of each error sub-model based on the error feature vector can be used in this invention.

[0148] The historical calibration samples include historical error feature vectors and corresponding error source labels. The historical error feature vectors are calculated from historical model predictions, historical reference measurements, and corresponding historical operating variables, historical environmental variables, historical motion variables, or historical process variables. The error source labels include one of the following: sensor measurement error labels, geometric error labels, thermal error labels, dynamic error labels, and process error labels.

[0149] The error source labels can be determined based on the known calibration item types, manual verification results, independent test results, or parameter rollback results from the historical calibration process. For example, zero-point offset samples confirmed in sensor zero-point measurement items are labeled as sensor measurement error labels; coordinate mapping deviation samples confirmed in multi-position reference measurement items are labeled as geometric error labels; thermal drift samples confirmed in no-load heating measurement items are labeled as thermal error labels; vibration response deviation samples confirmed in dynamic excitation measurement items are labeled as dynamic error labels; and load or material removal deviation samples confirmed in process parameter operation measurement items are labeled as process error labels.

[0150] S62: Based on the historical error feature vectors belonging to the same error source label, form the historical error feature template of the corresponding error sub-model, and determine the feature weights corresponding to the error sub-model;

[0151] In this embodiment, the feature weights can be represented in the form of a feature weight matrix.

[0152] For the k-th type of error sub-model, a historical error feature template corresponding to this type of error sub-model is formed based on historical calibration samples. and feature weight matrix They can be represented as:

[0153] (10)

[0154] (11)

[0155] in, This represents the number of historical calibration samples corresponding to the k-th type of error sub-model. Let represent the r-th historical error feature vector corresponding to the k-th type of error sub-model; , , , , and These represent the weights of the k-th error sub-model for error magnitude features, repeatability features, location-related features, temperature-related features, dynamic-related features, and process-related features, respectively.

[0156] The weights can be obtained from historical calibration samples or pre-set according to the physical mechanism corresponding to the error sub-model. For example, the thermal error sub-model can increase the weight of temperature-related features, the geometric error sub-model can increase the weight of position-related features, the dynamic error sub-model can increase the weight of dynamic-related features, and the process error sub-model can increase the weight of process-related features.

[0157] S63: Based on the current error feature vector, historical error feature templates, and feature weights, calculate the error source matching degree between the current error feature vector and each error sub-model;

[0158] In this embodiment, the error source matching degree between the current error feature vector and the k-th type error sub-model can be expressed as:

[0159] (12)

[0160] Where K represents the number of error sub-models, This represents the error source matching degree corresponding to the k-th error sub-model. The larger the error source matching degree, the closer the current error feature vector is to the historical error change pattern corresponding to the k-th error sub-model.

[0161] The above error source matching degree formula is an optional implementation method. Its function is to combine error source identification with manufacturing equipment error feature templates and feature weights, so as to avoid updating parameters solely based on the overall error magnitude.

[0162] S64: The error sub-model with the highest error source matching degree is determined as the error sub-model to be calibrated, and the error source matching degree or the value obtained based on the error source matching degree is used as the identification confidence degree;

[0163] In this embodiment, the error sub-model to be calibrated can be determined as follows:

[0164] (13)

[0165] The confidence level of identification can be expressed as:

[0166] (14)

[0167] in, The value represents the error sub-model number with the highest matching degree, and C represents the recognition confidence level.

[0168] S65: When the recognition confidence level is lower than the corresponding recognition confidence level threshold in the calibration strategy table, keep the parameters of the manufacturing equipment digital twin model unchanged and output a supplementary calibration or manual review prompt.

[0169] In this embodiment, when the recognition confidence level C is lower than the recognition confidence level threshold corresponding to the error sub-model to be calibrated... At that time, that is:

[0170] (15)

[0171] Keep the parameters of the digital twin model of the manufacturing equipment unchanged, and output supplementary calibration or manual verification prompts.

[0172] When the recognition confidence C reaches the recognition confidence threshold corresponding to the error sub-model to be calibrated... At that time, that is:

[0173] (16)

[0174] Select the corresponding preset calibration items based on the error sub-model to be calibrated, the identification confidence level, and the calibration strategy table.

[0175] S7: When the identification confidence level reaches the corresponding identification confidence level threshold in the calibration strategy table, select the corresponding preset calibration item according to the error sub-model to be calibrated, identification confidence level, error source matching degree and calibration strategy table, and control the manufacturing equipment to execute the preset calibration item; when the identification confidence level does not reach the corresponding identification confidence level threshold, keep the digital twin model parameters of the manufacturing equipment unchanged, and output a supplementary calibration or manual review prompt;

[0176] As attached Figure 5 As shown, in this embodiment, S7 includes the following steps:

[0177] S71: In the calibration strategy table, select preset calibration items that are consistent with the target error sub-model and the error sub-model to be calibrated;

[0178] In this embodiment, the error sub-model to be calibrated can be used as an index field to query the calibration strategy table, and the target error sub-model field of each preset calibration item can be read. Preset calibration items whose target error sub-model matches the error sub-model to be calibrated are then selected. This selection process ensures that subsequent calibration items first correspond to the source of error, avoiding the mixing of calibration items corresponding to thermal errors, geometric errors, sensor measurement errors, dynamic errors, or process errors.

[0179] S72: From the preset calibration items obtained by screening, retain the preset calibration items that have reached the corresponding recognition confidence threshold and have controlled variables or excitation variables and reference measurement objects that match the current operating status of the manufacturing equipment.

[0180] In this embodiment, "matching with the current operating state of the manufacturing equipment" means that the motion position, temperature range, speed range, load level, process parameters or measurement channels required for the preset calibration items can be satisfied by the current state of the manufacturing equipment, and the reference measurement object can be obtained in this state.

[0181] S73: Select the preset calibration item with the highest calibration priority from the reserved preset calibration items based on the calibration priority, error source matching degree, number of historical successful verifications, number of historical rollbacks, or the time of the most recent calibration item.

[0182] In this embodiment, if multiple preset calibration items are selected and retained, they can be sorted according to calibration priority. The calibration priority can be determined by a preset weight, error source matching degree, number of historical successful verifications, number of historical rollbacks, and the time since the most recent calibration. When calibration priorities are the same or close, preset calibration items with higher error source matching degree, fewer historical rollbacks, or a longer time since the most recent calibration can be prioritized. This sorting and selection process ensures that the selected preset calibration items balance the current error source matching degree and historical execution reliability.

[0183] S74: When there is no preset calibration item with an identification confidence level that reaches the corresponding identification confidence level threshold, or when there is no preset calibration item that matches the current operating status of the manufacturing equipment, keep the parameters of the digital twin model of the manufacturing equipment unchanged and output a prompt for supplementary calibration or manual review.

[0184] In this embodiment, the absence of a pre-defined calibration item that meets the conditions includes two situations: First, the identification confidence level corresponding to the error sub-model to be calibrated does not reach the identification confidence level threshold corresponding to the calibration strategy table, indicating that the error source identification result is insufficient to support automatic calibration; second, although there is a pre-defined calibration item consistent with the target error sub-model, the current manufacturing equipment's motion position, temperature range, speed range, load level, process parameters, measurement channel, or reference measurement object cannot meet the execution conditions of the pre-defined calibration item. When the above situations occur, the system does not generate candidate update quantities, does not change the digital twin model parameters, and records the reason for non-execution for subsequent supplementation of calibration data or manual review.

[0185] S8: During the process of the manufacturing equipment performing the preset calibration items, the actual values ​​of the controlled variable or excitation variable and the reference measurement values ​​corresponding to the reference measurement object are obtained, and the calibration input variables, including the actual values ​​of the controlled variable or excitation variable, are input into the digital twin model of the manufacturing equipment to obtain the model prediction value corresponding to the reference measurement value, and the calibration error between the model prediction value and the reference measurement value is calculated.

[0186] In this embodiment, the preset calibration items, controlled variables or excitation variables, and reference measurement values ​​corresponding to different error sub-models can be set as follows.

[0187] When the error sub-model to be calibrated is a sensor measurement error sub-model, the preset calibration items may include sensor zero-point measurement items, repeated measurement items at the same position on a fixed reference surface, or standard value input measurement items. The controlled variable or excitation variable includes the standard input value, the number of repeated measurements, or the sensor zero-point state. The reference measurement value includes the sensor zero-point measurement value, the repeated measurement value on a fixed reference surface, or the sensor output value under standard value input conditions.

[0188] When the error sub-model to be calibrated is a geometric error sub-model, the preset calibration items can include multi-position reference measurement items, standard trajectory scanning items, or bidirectional reciprocating positioning measurement items. The controlled or excitation variables include the position of the motion actuator, the direction of motion, the trajectory point number, or the direction of reciprocating motion. Reference measurement values ​​include reference component measurement values, standard trajectory measurement values, or multi-position geometric measurement values.

[0189] When the error sub-model to be calibrated is a thermal error sub-model, the preset calibration items can include no-load temperature rise measurement items, constant-rate thermal stability measurement items, or cooling process measurement items. The controlled variable or excitation variable includes temperature, heat source power, running time, or cooling time. Reference measurements include thermal drift measurements or displacement measurements under temperature change conditions.

[0190] When the error sub-model to be calibrated is a dynamic error sub-model, the preset calibration items can include variable speed motion measurement items, step motion measurement items, reciprocating acceleration / deceleration measurement items, or dynamic excitation measurement items. Controlled or excitation variables include rotational speed, velocity, acceleration, excitation frequency, or motion reversal state. Reference measurements include vibration response measurements, dynamic displacement measurements, or motion response measurements.

[0191] When the error sub-model to be calibrated is a process error sub-model, the preset calibration items can include measurement items based on process parameters, load levels, or energy inputs. Controlled or excitation variables include process parameters, load levels, power, current, or contact conditions. Reference measurements include process load measurements, post-processed manufactured object measurements, or material removal measurements.

[0192] When controlling manufacturing equipment to perform preset calibration items, calibration execution instructions can be issued to the CNC system, motion controller, temperature control unit, excitation source, or data acquisition unit of the manufacturing equipment according to the calibration execution method in the calibration strategy table. The calibration execution instructions include one or more of the following: motion trajectory, measurement position, running speed, rotational speed, acceleration, temperature holding time, excitation frequency, sampling channel, sampling frequency, and number of repeated measurements.

[0193] During the process of performing a preset calibration project on the manufacturing equipment, the actual values ​​of the controlled variable or excitation variable and the reference measurement value corresponding to the reference measurement object are obtained. The calibration input variables, including the actual values ​​of the controlled variable or excitation variable, are input into the digital twin model of the manufacturing equipment to obtain the model prediction value corresponding to the reference measurement value. The calibration error between the model prediction value and the reference measurement value is calculated.

[0194] After the manufacturing equipment performs the preset calibration items, the system records the set value, actual value, execution time and reference measurement value of the controlled variable or excitation variable, which are used for subsequent calibration error calculation and intelligent calibration condition judgment.

[0195] S9: Calculate the intelligent calibration confidence level based on the identification confidence level, error source matching degree, target error sub-model corresponding to the preset calibration item, excitation coverage of the controlled variable or excitation variable, repeatability of the reference measurement value, error mode matching factor and calibration error, and determine whether the preset calibration item meets the intelligent calibration conditions based on the intelligent calibration confidence level.

[0196] In this embodiment, S9 includes the following steps:

[0197] S91: Determine whether the target error sub-model corresponding to the preset calibration item is consistent with the error sub-model to be calibrated;

[0198] In this embodiment, if the target error sub-model is consistent with the error sub-model to be calibrated, it indicates that the preset calibration item corresponds to the current object to be calibrated at the error source level, and subsequent judgments such as confidence, excitation coverage, and repeatability can continue. If the two are inconsistent, even if the preset calibration item can be executed, it will not be used as the basis for updating the current parameters. This avoids using deviations caused by non-target error sources to update the parameters corresponding to the current error sub-model to be calibrated.

[0199] S92: Determine whether the recognition confidence level reaches the recognition confidence threshold corresponding to the preset calibration item;

[0200] In this embodiment, different recognition confidence thresholds can be set for different error sub-models or different preset calibration items. For example, a higher threshold can be set for sensor zero-point measurement items to reduce misjudgments caused by random noise; for thermal error calibration items that require a long-term thermal stabilization process, the threshold can be set in conjunction with the stability of historical samples. Only when the recognition confidence reaches the corresponding threshold is the error source recognition result considered sufficient to support the subsequent intelligent calibration confidence calculation.

[0201] S93: Determine whether the actual value of the controlled variable or the excitation variable reaches the preset range of change corresponding to the target error sub-model, and determine the excitation coverage based on the relationship between the actual value and the preset range of change;

[0202] In this embodiment, for the k-th type of error sub-model, let... This indicates the actual range of change of the controlled or excitation variable during the execution of the preset calibration project. Let represent the minimum effective range of variation corresponding to this error sub-model. Then, the excitation coverage can be expressed as:

[0203] (17)

[0204] in, Indicates incentive coverage. This indicates a positive number used to avoid a denominator of zero. When there are multiple controlled or excitation variables, It can be the weighted sum, maximum or minimum value of the normalized range of each controlled variable or stimulus variable; the weight of each controlled variable or stimulus variable is determined by the target error sub-model or preset calibration items.

[0205] S94: Determine whether the repeatability of the reference measurement value meets the preset repeatability requirement, and determine the reference measurement repeatability factor based on the dispersion of the reference measurement value;

[0206] In this embodiment, the reference measurement repeatability factor can be expressed as:

[0207] (18)

[0208] in, Indicates the repeatability factor of the reference measurement. This represents the repeatability threshold corresponding to the k-th type of error sub-model. The greater the dispersion of the reference measurement, the lower the repeatability threshold. The smaller.

[0209] S95: Determine whether the calibration error exceeds a preset error threshold, and determine the error amplitude validity factor based on the relationship between the calibration error and the preset error threshold;

[0210] In this embodiment, the error amplitude validity factor can be expressed as:

[0211] (19)

[0212] in, Indicates the validity factor of the error magnitude. This represents the error threshold corresponding to the k-th type of error sub-model.

[0213] S96: Determine whether the characteristics of the calibration error changing with the controlled variable or the excitation variable match the error change characteristics corresponding to the target error sub-model, and determine the error mode matching factor based on the degree of matching of the error source or the degree of consistency of the error change characteristics;

[0214] In this embodiment, the error pattern matching factor is used to characterize the consistency between the calibration error change pattern obtained during the execution of the preset calibration project and the corresponding error change characteristics of the target error sub-model. It is different from the error source matching degree calculated based on the initial deviation in the error source identification stage.

[0215] The error pattern matching factor can be expressed as:

[0216] (20)

[0217] in, The error pattern matching factor is used to characterize the consistency between the calibration error variation pattern obtained during the execution of the preset calibration project and the corresponding error variation characteristics of the target error sub-model.

[0218] S97: Determine the confidence level of intelligent calibration based on the identification confidence level, excitation coverage, reference measurement repeatability factor, error pattern matching factor, and error amplitude validity factor;

[0219] In this embodiment, the reliability of intelligent calibration can be expressed as:

[0220] (twenty one)

[0221] in, This represents the intelligent calibration reliability of the preset calibration item corresponding to the k-th type of error sub-model. This represents the recognition confidence level corresponding to the k-th type of error sub-model. When the k-th type of error sub-model is the error sub-model to be calibrated, Equals the identification confidence level C. When:

[0222] (twenty two)

[0223] Furthermore, if the target error sub-model corresponding to the preset calibration item is consistent with the error sub-model to be calibrated, the preset calibration item is determined to meet the intelligent calibration conditions. This represents the intelligent calibration confidence threshold corresponding to the k-th type of error sub-model.

[0224] By using the aforementioned intelligent calibration confidence assessment, we can avoid updating digital twin model parameters when there is insufficient confidence, insufficient excitation range, poor repeatability of reference measurements, mismatch in error variation characteristics, or insufficient calibration error amplitude, thereby reducing the risk of erroneous parameter updates. For example, when thermal drift error is significantly correlated with temperature variables but weakly correlated with location variables, the error mode matching factor of the thermal error sub-model is high; when the same deviation mainly changes with location and is independent of temperature changes, the error mode matching factor of the geometric error sub-model is high. Therefore, we can reduce the risk of incorrectly updating thermal error parameters to geometric mapping parameters and vice versa.

[0225] S10: When the preset calibration item meets the intelligent calibration conditions, the candidate update amount of the digital twin model parameters within the allowed update parameter set, parameter sensitivity matrix, and parameter update constraints in the preset calibration item is calculated, and the digital twin model parameters that do not belong to the allowed update parameter set are kept unchanged to obtain the candidate digital twin model parameters; when the preset calibration item does not meet the intelligent calibration conditions, the digital twin model parameters of the manufacturing equipment are kept unchanged.

[0226] In this embodiment, S10 includes the following steps:

[0227] S101: Determine the set of parameters that can be updated in the calibration strategy table based on the target error sub-model corresponding to the preset calibration item;

[0228] In this embodiment, the set of parameters that can be updated is denoted as A. When determining the set of parameters that can be updated, the system only reads the parameter fields associated with the target error sub-model and marks the parameters corresponding to other error sub-models as non-updateable parameters. For example, when the target error sub-model is a sensor measurement error sub-model, the set of parameters that can be updated may only include sensor zero-point parameters, sensitivity parameters, range correction parameters, or measurement noise parameters; when the target error sub-model is a thermal error sub-model, the set of parameters that can be updated may only include temperature sensitivity coefficients, thermal hysteresis time, thermal drift mapping coefficients, or thermal boundary correction coefficients. By limiting the set, subsequent sensitivity matrix calculations and candidate update quantity calculations are only performed on parameters within the set.

[0229] S102: Based on the actual values ​​of the controlled or excitation variables collected during the calibration process, calculate the local sensitivity of the model prediction values ​​to each parameter within the allowed update parameter set, and obtain the parameter sensitivity matrix;

[0230] In this embodiment, before updating the parameters of the digital twin model within the allowed update parameter set, the local sensitivity of the model prediction values ​​to each parameter in the allowed update parameter set is calculated based on the actual values ​​of the controlled or excitation variables collected during the calibration process, thus obtaining the parameter sensitivity matrix. , can be represented as:

[0231] (twenty three)

[0232] in, Represents the vector of model predicted values. This indicates that updating the parameter vector within parameter set A is allowed.

[0233] S103: Based on the calibration error and parameter sensitivity matrix, calculate the candidate update amounts for the parameters of the digital twin model within the allowed parameter set;

[0234] In this embodiment, the candidate update amount can be obtained using a constrained weighted sensitivity update method:

[0235] (twenty four)

[0236] in, This represents the calibration error vector composed of the deviations at each sampling point during the calibration process. Represents the error weight matrix. This indicates the parameter update constraint weight matrix. Represents the regularization coefficient. This represents the vector of the maximum single update amount allowed for each parameter within the parameter set. This represents an element-wise truncation function, used to limit the candidate update amounts of each parameter in the allowed update parameter set to the corresponding maximum single update amount range.

[0237] S104: Determine whether the candidate update quantity meets the requirements of maximum single update quantity, minimum update interval, upper limit of cumulative change, and parameter value range based on the parameter update constraints;

[0238] In this embodiment, for any parameter in the set of parameters A that can be updated... The following constraints can be adopted:

[0239] (25)

[0240] (26)

[0241] in, Indicates parameters The number of candidate updates, Indicates parameters The maximum single update size, and Representing parameters respectively The lower limit and upper limit of the value, Indicates parameters Updated candidate values.

[0242] S105: When the candidate update quantity satisfies the parameter update constraint, candidate digital twin model parameters are generated; when the candidate update quantity does not satisfy the parameter update constraint, the candidate update quantity is not enabled, and the digital twin model parameters before the update remain unchanged.

[0243] In this embodiment, the parameters of the manufacturing equipment digital twin model before the update are assumed to be: The number of candidate updates is Then the parameters of the candidate digital twin model can be expressed as:

[0244] (27)

[0245] in, and Let A represent the candidate digital twin model parameters that belong to the set of allowed update parameters, and the parameters before the update, respectively. For digital twin model parameters that do not belong to the set of allowed update parameters A, their parameter values ​​are kept unchanged and can be represented as follows:

[0246] (28)

[0247] in, This indicates a set of parameters that are not allowed to be updated in parameter set A. and These represent the candidate digital twin model parameters that are not included in the set of allowed update parameters, and the parameters before the update, respectively.

[0248] Specifically, when the preset calibration item corresponds to the sensor measurement error sub-model, the parameter set that can be updated includes sensor zero-point parameters, sensitivity parameters, range correction parameters, or measurement noise parameters.

[0249] When the preset calibration item corresponds to the geometric error sub-model, the parameter set that can be updated includes coordinate mapping parameters, installation pose parameters, measurement direction parameters, motion axis positioning error parameters, straightness error parameters, or backlash parameters;

[0250] When the preset calibration item corresponds to the thermal error sub-model, the set of parameters that can be updated includes the temperature sensitivity coefficient, thermal hysteresis time, thermal drift mapping coefficient, or thermal boundary correction coefficient.

[0251] When the preset calibration item corresponds to the dynamic error sub-model, the parameter set that can be updated includes dynamic stiffness parameters, damping parameters, vibration transmission parameters, speed-related response parameters, or acceleration / deceleration response parameters.

[0252] When the preset calibration item corresponds to the process error sub-model, the parameter set that can be updated includes process load mapping parameters, contact stiffness parameters, power mapping parameters, current mapping parameters, or material removal effect parameters.

[0253] When a candidate update value satisfies the parameter update constraints, candidate digital twin model parameters are generated; when a candidate update value does not satisfy the parameter update constraints, it is not used, and the digital twin model parameters before the update remain unchanged. For example, when the identification result is a sensor measurement error sub-model, only the sensor zero-point parameters, sensitivity parameters, range correction parameters, or measurement noise parameters are allowed to be updated, while geometric mapping parameters, thermal drift mapping coefficients, dynamic stiffness parameters, or process load mapping parameters are not updated. This reduces the risk of incorrect parameter updates caused by error source confusion.

[0254] S11: After obtaining the candidate digital twin model parameters, execute the verification project, obtain the reference measurement value corresponding to the verification project, and calculate the updated verification error and the original verification error based on the candidate digital twin model parameters and the original digital twin model parameters, respectively. Then, determine the verification benefit index based on the updated verification error, the original verification error, and the candidate update amount.

[0255] In this embodiment, S11 includes the following steps:

[0256] S111: After obtaining the parameters of the candidate digital twin model, a verification item is selected. The verification item is different from the preset calibration item used for parameter updating; or, the verification item and the preset calibration item used for parameter updating belong to the same type of preset calibration item, but their measurement positions, motion trajectories, running speeds, temperature ranges, load levels, process parameters, or running time periods are different.

[0257] As attached Figure 6 As shown in this embodiment, after obtaining the parameters of the candidate digital twin model, a verification project is performed. By setting verification projects that differ from the calibration projects, the evaluation of the candidate digital twin model parameters can be avoided by directly using the data used for parameter updates.

[0258] S112: Perform the verification item and obtain the reference measurement value corresponding to the verification item;

[0259] In this embodiment, when executing a verification project, the system controls the manufacturing equipment and data acquisition unit to operate according to the measurement position, motion trajectory, running speed, temperature range, load level, process parameters, or running time period corresponding to the verification project, and obtains the reference measurement value corresponding to the verification project. The reference measurement value can be obtained from an independent measuring device, standard value input, reference part measurement, post-processed manufactured object detection, or a measurement position different from the calibration project.

[0260] S113: Calculate the updated verification error and the original verification error based on the candidate digital twin model parameters and the original digital twin model parameters, respectively;

[0261] In this embodiment, the updated verification error can be expressed as:

[0262] (29)

[0263] The verification error before the update can be expressed as:

[0264] (30)

[0265] in, This indicates the updated verification error. This indicates the verification error before the update. This indicates the number of sampling points in the verification project. This represents the reference measurement value of the i-th sampling point in the verification project. This represents the model prediction value at the i-th sampling point when using candidate digital twin model parameters. This represents the model prediction value of the i-th sampling point when using the parameters of the digital twin model before the update.

[0266] S114: Determine the verification benefit index corresponding to the candidate digital twin model parameters based on the updated verification error, the original verification error, and the candidate update amount within the allowed update parameter set;

[0267] In this embodiment, the verification return index can be expressed as:

[0268] (31)

[0269] Where B represents the validation return index, This represents the penalty coefficient for parameter changes. Let L2 represent the norm. The first term in the formula represents the relative reduction in verification error, and the vector division in the second term represents the element-wise ratio of each candidate update amount in the allowed update parameter set to the corresponding maximum single update amount. This term is used to penalize excessively large parameter update amounts.

[0270] S12: Determine the parameter processing method based on the verification benefit index, the updated verification error, the verification error before the update, the parameter update constraints, and the historical verification results. Save and enable the candidate digital twin model parameters or roll back to the digital twin model parameters before the update according to the parameter processing method, and update the calibration priority of the preset calibration items.

[0271] In this embodiment, S12 includes the following steps:

[0272] S121: Determine whether the verification benefit index has reached the preset verification benefit threshold;

[0273] In this embodiment, when the following conditions are met:

[0274] (32)

[0275] When this occurs, it indicates that the validation benefit of the candidate digital twin model parameters has reached a preset validation benefit threshold. Among these, This indicates a preset verification benefit threshold. When:

[0276] (33)

[0277] When the verification benefit of the candidate digital twin model parameters does not reach the preset verification benefit threshold, it indicates that the verification benefit of the candidate digital twin model parameters has not reached the preset verification benefit threshold.

[0278] S122: Based on the parameter update constraints, determine whether the parameters of the candidate digital twin model meet the requirements of maximum single update amount, minimum update interval, upper limit of cumulative change amount, or parameter value range.

[0279] In this embodiment, for each parameter within the allowed update parameter set, it is determined whether its candidate update amount exceeds the maximum single update amount, whether the time or running cycle since the parameter's last update meets the minimum update interval, whether the parameter's cumulative change exceeds the upper limit of cumulative change, and whether the updated parameter value falls within the preset parameter value range. If any of the above constraints are not met, it can be determined that the candidate digital twin model parameter does not meet the parameter update constraints.

[0280] S123: Based on historical verification results, determine whether the same model parameters, the same error sub-model, or the same preset calibration item have reached the preset historical rollback limit;

[0281] In this embodiment, historical verification results may include the number of successful historical calibrations, the number of historical rollbacks, the reasons for historical rollbacks, the historical verification benefit index, the amount of historical parameter updates, and the corresponding error sub-models, preset calibration items, and model parameter identifiers. When determining the historical rollback limit, the same model parameter, the same error sub-model, or the same preset calibration item can be used as the statistical object to count the number of rollbacks within a preset time window, a preset running cycle, or a preset number of calibrations; when the number of rollbacks reaches the preset number, it is considered that the preset historical rollback limit has been reached.

[0282] S124: When the verification benefit index reaches the preset verification benefit threshold, the candidate digital twin model parameters meet the parameter update constraints, and the preset historical rollback limit is not reached, the parameter processing method is determined to be save and enable, the candidate digital twin model parameters are saved and enabled, and a new digital twin model version is generated.

[0283] In this embodiment, when the judgment results of S121 to S123 all support saving and enabling, the system writes the candidate digital twin model parameters into the model parameter library or model version library and generates a new digital twin model version. The new digital twin model version can record the version number, update time, corresponding preset calibration items, verification items, verification benefit index, candidate update quantity, and parameter values ​​before the update. After saving and enabling, subsequent model prediction, state evaluation, or compensation control will call the new digital twin model parameters and record a historical calibration success result for subsequent updates to calibration priority.

[0284] S125: When the verification benefit index does not reach the preset verification benefit threshold, or the candidate digital twin model parameters do not meet the parameter update constraints, or the preset historical rollback limit is reached, the parameter processing method is determined to be rollback, rollback to the digital twin model parameters before the update, and the rollback reason is recorded.

[0285] In this embodiment, when the verification benefit index fails to reach the preset verification benefit threshold, the candidate digital twin model parameters do not meet the parameter update constraints, or the same model parameter, the same error sub-model, or the same preset calibration item reaches the preset historical rollback limit, the system will not activate the candidate digital twin model parameters and will restore the model parameters to their pre-update state. The rollback reasons may include insufficient verification benefit, exceeding the single update limit, parameter value exceeding the limit, exceeding the cumulative change limit, insufficient minimum update interval, or exceeding the historical rollback limit. Rollback records are used to subsequently reduce the calibration priority of the corresponding preset calibration item and provide a basis for manual review or recalibration.

[0286] S126: Update the calibration priority of the preset calibration item according to the parameter processing method, wherein when the parameter processing method is save enabled, increase the calibration priority of the corresponding preset calibration item or record a historical calibration success count; when the parameter processing method is rollback, decrease the calibration priority of the corresponding preset calibration item or record a historical rollback count.

[0287] In this embodiment, when verification is successful, the calibration priority can be increased in the following way:

[0288] (34)

[0289] When verification fails, the calibration priority can be reduced in the following ways:

[0290] (35)

[0291] in, Indicates the calibration priority before the update. Indicates the updated calibration priority. and These represent the upper and lower limits of calibration priority, respectively. Indicates the success reward coefficient. This represents the failure penalty coefficient. The priority update method described above is only an example; other methods can also be used to determine calibration priority based on the number of historical successful calibrations, the number of historical rollbacks, the time of the most recent calibration, or the change in verification error.

[0292] S127: When the number of historical rollbacks for the same model parameter, the same error sub-model, or the same preset calibration item reaches the preset number, pause the automatic parameter update of the corresponding model parameter, error sub-model, or preset calibration item, and output a prompt for recalibration or manual review.

[0293] In this embodiment, when the number of historical rollbacks for the same model parameter reaches a preset number, the model parameter is marked as paused for automatic updates; when the number of historical rollbacks for the same error sub-model reaches a preset number, the error sub-model is marked as requiring manual review; when the number of historical rollbacks for the same preset calibration item reaches a preset number, the calibration priority of the preset calibration item is reduced or frozen, and the preset calibration item is skipped during subsequent calibration item screening until recalibration or manual review is completed.

[0294] The recalibration or manual review prompts may include the object type that has reached the historical rollback limit, the object name, the number of historical rollbacks, the reason for the most recent rollback, the corresponding error sub-model, the corresponding preset calibration item, and the suggested recalibration item. This process avoids the same parameter, the same error sub-model, or the same preset calibration item from being repeatedly and automatically updated after consecutive verification failures, thereby further reducing the risk of parameter drift and repeated erroneous updates in the digital twin model.

[0295] Through the above-mentioned verification, save and enable, rollback and priority feedback mechanisms, it is possible to avoid the situation where candidate parameters are only valid on the original calibration data and fail under other measurement locations, operating states, temperature ranges, load levels or process parameters, thereby improving the stability and reliability of the long-term operation of the digital twin model of manufacturing equipment.

[0296] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Those skilled in the art can make various changes and modifications without departing from the concept of the present invention, and all such changes and modifications should fall within the protection scope of the present invention.

Claims

1. A method for intelligent calibration and parameter updating of a digital twin model of manufacturing equipment, characterized in that, include: S1: Establish a digital twin model of manufacturing equipment. The digital twin model of manufacturing equipment is used to output model prediction values ​​corresponding to the state of manufacturing equipment or the state of manufacturing object based on the operating variables, environmental variables, motion variables or process variables of the manufacturing equipment. S2: Based on the impact of different error sources on the model's predicted values, the digital twin model of the manufacturing equipment is divided into several error sub-models; The error sub-model includes two or more of the following: sensor measurement error sub-model, geometric error sub-model, thermal error sub-model, dynamic error sub-model, and process error sub-model. The sensor measurement error sub-model characterizes the influence of sensor zero-point offset, sensitivity variation, range deviation, or measurement noise on the model's predicted values. The geometric error sub-model characterizes the influence of coordinate mapping errors, installation posture errors, or motion geometric errors between the equipment coordinate system, sensor coordinate system, motion actuator coordinate system, and manufacturing object coordinate system on the model's predicted values. The thermal error sub-model characterizes the influence of temperature changes, heat source power changes, thermal hysteresis, or thermal deformation on the model's predicted values. The dynamic error sub-model characterizes the influence of changes in rotational speed, velocity, acceleration, vibration, damping, or dynamic stiffness on the model's predicted values. The process error sub-model characterizes the forces during cutting, grinding, polishing, or surface treatment processes. The impact of power, current, contact state, or material removal on model predictions; S3: Establish a calibration strategy table, which is used to record the correspondence between error sub-model, preset calibration items, allowed update parameter set, parameter update constraints, verification items, historical error feature template, feature weight, identification confidence threshold, intelligent calibration confidence threshold and calibration priority; S4: Obtain the deviation between the model prediction value and the reference measurement value of the digital twin model of the manufacturing equipment; S5: Extract the error feature vector based on the deviation and the repeatability of the corresponding operating variables, environmental variables, motion variables, process variables or reference measurement values; The error feature vector includes two or more of the following features: Error magnitude characteristics are used to characterize the magnitude of the deviation between the model's predicted values ​​and the reference measured values; Repeatability characteristics are used to characterize the degree of dispersion of reference measurement values ​​at the same measurement location, under the same standard input value, or under the same operating conditions. Position-related features are used to characterize the degree to which deviation changes with measurement position, direction of motion, trajectory point number, or reciprocating motion direction; Temperature-dependent characteristics are used to characterize the degree to which deviations vary with temperature, heat source power, operating time, or cooling time. Dynamic correlation characteristics are used to characterize the degree to which deviation changes with rotational speed, velocity, acceleration, excitation frequency, or motion reversal state; Process-related characteristics are used to characterize the extent to which deviations vary with process parameters, load levels, power, current, or contact conditions. Furthermore, the historical error feature template is determined based on the historical error feature vector in the historical calibration samples, and the feature weight is determined based on the statistical results of the historical calibration samples or the physical mechanism corresponding to each error sub-model; S6: Based on the error feature vector and the historical error feature template and feature weight corresponding to each error sub-model, calculate the error source matching degree of each error sub-model, and obtain the error sub-model to be calibrated and the identification confidence level corresponding to the error sub-model to be calibrated; S7: When the identification confidence level reaches the corresponding identification confidence level threshold in the calibration strategy table, select the corresponding preset calibration item according to the error sub-model to be calibrated, identification confidence level, error source matching degree and calibration strategy table, and control the manufacturing equipment to execute the preset calibration item; when the identification confidence level does not reach the corresponding identification confidence level threshold, keep the digital twin model parameters of the manufacturing equipment unchanged, and output a supplementary calibration or manual review prompt; S8: During the process of the manufacturing equipment performing the preset calibration items, the actual values ​​of the controlled variable or excitation variable and the reference measurement values ​​corresponding to the reference measurement object are obtained, and the calibration input variables, including the actual values ​​of the controlled variable or excitation variable, are input into the digital twin model of the manufacturing equipment to obtain the model prediction value corresponding to the reference measurement value, and the calibration error between the model prediction value and the reference measurement value is calculated. S9: Calculate the intelligent calibration confidence level based on the identification confidence level, error source matching degree, target error sub-model corresponding to the preset calibration item, excitation coverage of the controlled variable or excitation variable, repeatability of the reference measurement value, error mode matching factor and calibration error, and determine whether the preset calibration item meets the intelligent calibration conditions based on the intelligent calibration confidence level. S10: When the preset calibration item meets the intelligent calibration conditions, the candidate update amount of the digital twin model parameters within the allowed update parameter set, parameter sensitivity matrix, and parameter update constraints in the preset calibration item is calculated, and the digital twin model parameters that do not belong to the allowed update parameter set are kept unchanged to obtain the candidate digital twin model parameters; when the preset calibration item does not meet the intelligent calibration conditions, the digital twin model parameters of the manufacturing equipment are kept unchanged. S11: After obtaining the candidate digital twin model parameters, execute the verification project, obtain the reference measurement value corresponding to the verification project, and calculate the updated verification error and the original verification error based on the candidate digital twin model parameters and the original digital twin model parameters, respectively. Then, determine the verification benefit index based on the updated verification error, the original verification error, and the candidate update amount. S12: Determine the parameter processing method based on the verification benefit index, the updated verification error, the verification error before the update, the parameter update constraints, and the historical verification results. Save and enable the candidate digital twin model parameters or roll back to the digital twin model parameters before the update according to the parameter processing method, and update the calibration priority of the preset calibration items.

2. The intelligent calibration and parameter update method for digital twin models of manufacturing equipment as described in claim 1, characterized in that, The calibration strategy table in step S3 includes several preset calibration items. Each preset calibration item corresponds to the target error sub-model, calibration execution method, controlled variable or excitation variable, reference measurement object, set of parameters that can be updated, parameter update constraints, verification items, historical error feature template, feature weight, identification confidence threshold, intelligent calibration confidence threshold, and calibration priority. The target error sub-model is used to represent the error source corresponding to the preset calibration item; The calibration execution mode is used to indicate the motion mode, operation mode, or process execution mode of the manufacturing equipment when performing the preset calibration item; The controlled variable or excitation variable is used to represent the variable that is actively changed or controlled to be maintained when the preset calibration item is performed; The reference measurement object is used to compare with the model prediction value output by the digital twin model of the manufacturing equipment; The set of allowed update parameters is used to limit the digital twin model parameters that can be updated when the preset calibration item meets the intelligent calibration conditions; The parameter update constraint is used to limit the maximum single update amount, minimum update interval, upper limit of cumulative change, or range of parameter values ​​within the parameter set that are allowed to be updated. The historical error feature template and feature weights are used to calculate the error source matching degree between the current error feature vector and each error sub-model; The verification project is used to verify the parameters of the candidate digital twin model after the parameters are updated; The calibration priority is determined based on the importance of the error sub-model corresponding to the preset calibration item, the number of successful historical calibrations, the number of historical rollbacks, the time of the most recent calibration, or the verification benefit results.

3. The intelligent calibration and parameter update method for digital twin models of manufacturing equipment as described in claim 1, characterized in that, In step S6, obtaining the error sub-model to be calibrated and identifying the confidence level includes: S61: Obtain historical calibration samples, which include historical error feature vectors and corresponding error source labels. The error source labels include one of the following: sensor measurement error label, geometric error label, thermal error label, dynamic error label, and process error label. S62: Based on the historical error feature vectors belonging to the same error source label, form the historical error feature template of the corresponding error sub-model, and determine the feature weights corresponding to the error sub-model; S63: Based on the current error feature vector, historical error feature templates, and feature weights, calculate the error source matching degree between the current error feature vector and each error sub-model; S64: The error sub-model with the highest error source matching degree is identified as the error sub-model to be calibrated, and the error source matching degree or the value obtained based on the error source matching degree is used as the identification confidence degree; S65: When the recognition confidence level is lower than the corresponding recognition confidence level threshold in the calibration strategy table, keep the parameters of the manufacturing equipment digital twin model unchanged and output a supplementary calibration or manual review prompt.

4. The intelligent calibration and parameter update method for digital twin models of manufacturing equipment as described in claim 1, characterized in that, In step S7, selecting the corresponding preset calibration items based on the error sub-model to be calibrated, the identification confidence level, the error source matching degree, and the calibration strategy table includes: S71: In the calibration strategy table, select preset calibration items that are consistent with the target error sub-model and the error sub-model to be calibrated; S72: From the preset calibration items obtained by screening, retain the preset calibration items that have reached the corresponding recognition confidence threshold and have controlled variables or excitation variables and reference measurement objects that match the current operating status of the manufacturing equipment. S73: Select the preset calibration item with the highest calibration priority from the reserved preset calibration items based on the calibration priority, error source matching degree, number of historical successful verifications, number of historical rollbacks, or the time of the most recent calibration item. S74: When there is no preset calibration item with an identification confidence level that reaches the corresponding identification confidence level threshold, or when there is no preset calibration item that matches the current operating status of the manufacturing equipment, keep the parameters of the digital twin model of the manufacturing equipment unchanged and output a prompt for supplementary calibration or manual review.

5. The intelligent calibration and parameter update method for digital twin models of manufacturing equipment as described in claim 1, characterized in that, In step S8, the preset calibration items, controlled variables or excitation variables, and reference measurement values ​​corresponding to different error sub-models include: When the preset calibration items correspond to the sensor measurement error sub-model, the preset calibration items include sensor zero-position measurement items, repeated measurement items at the same position on a fixed reference surface, or standard value input measurement items; the controlled variable or excitation variable includes standard input value, number of repeated measurements, or sensor zero-position state; the reference measurement value includes sensor zero-position measurement value, repeated measurement value on a fixed reference surface, or sensor output value under standard value input conditions; When the preset calibration items correspond to the geometric error sub-model, the preset calibration items include multi-position reference measurement items, standard trajectory scanning items, or bidirectional reciprocating positioning measurement items; the controlled variables or excitation variables include the position of the motion actuator, the direction of motion, the trajectory point number, or the direction of reciprocating motion; the reference measurement values ​​include reference component measurement values, standard trajectory measurement values, or multi-position geometric measurement values; When the preset calibration items correspond to the thermal error sub-model, the preset calibration items include no-load heating measurement items, constant-rate thermal stability measurement items, or cooling process measurement items; the controlled variable or excitation variable includes temperature, heat source power, running time, or cooling time; the reference measurement value includes thermal drift measurement value or displacement measurement value under temperature change conditions; When the preset calibration items correspond to the dynamic error sub-model, the preset calibration items include variable speed motion measurement items, step motion measurement items, reciprocating acceleration and deceleration measurement items, or dynamic excitation measurement items; the controlled variable or excitation variable includes rotational speed, velocity, acceleration, excitation frequency, or motion reversal state; the reference measurement values ​​include vibration response measurement values, dynamic displacement measurement values, or motion response measurement values. When the preset calibration items correspond to the process error sub-model, the preset calibration items include process parameter operation measurement items, load level operation measurement items, or energy input operation measurement items; the controlled variables or excitation variables include process parameters, load levels, power, current, or contact status; the reference measurement values ​​include process load measurement values, post-processed manufactured object measurement values, or material removal measurement values.

6. The intelligent calibration and parameter update method for digital twin models of manufacturing equipment as described in claim 1, characterized in that, In step S9, determining whether the preset calibration item meets the intelligent calibration conditions includes: S91: Determine whether the target error sub-model corresponding to the preset calibration item is consistent with the error sub-model to be calibrated; S92: Determine whether the recognition confidence level reaches the recognition confidence threshold corresponding to the preset calibration item; S93: Determine whether the actual value of the controlled variable or the excitation variable reaches the preset range of change corresponding to the target error sub-model, and determine the excitation coverage based on the relationship between the actual value and the preset range of change; S94: Determine whether the repeatability of the reference measurement value meets the preset repeatability requirement, and determine the reference measurement repeatability factor based on the dispersion of the reference measurement value; S95: Determine whether the calibration error exceeds a preset error threshold, and determine the error amplitude validity factor based on the relationship between the calibration error and the preset error threshold; S96: Determine whether the characteristics of the calibration error changing with the controlled variable or the excitation variable match the error change characteristics corresponding to the target error sub-model, and determine the error mode matching factor based on the degree of matching of the error source or the degree of consistency of the error change characteristics; S97: Determine the confidence level of intelligent calibration based on the identification confidence level, excitation coverage, reference measurement repeatability factor, error pattern matching factor, and error amplitude validity factor; When the target error sub-model corresponding to the preset calibration item is consistent with the error sub-model to be calibrated, and the smart calibration confidence level reaches the corresponding smart calibration confidence level threshold, the preset calibration item is determined to meet the smart calibration conditions; when the target error sub-model corresponding to the preset calibration item is inconsistent with the error sub-model to be calibrated, or the smart calibration confidence level does not reach the corresponding smart calibration confidence level threshold, the preset calibration item is determined to not meet the smart calibration conditions.

7. The intelligent calibration and parameter update method for digital twin models of manufacturing equipment as described in claim 1, characterized in that, In step S10, updating only the digital twin model parameters within the set of allowed update parameters in the preset calibration project includes: S101: Determine the set of parameters that can be updated in the calibration strategy table based on the target error sub-model corresponding to the preset calibration item; S102: Based on the actual values ​​of the controlled or excitation variables collected during the calibration process, calculate the local sensitivity of the model prediction values ​​to each parameter within the allowed update parameter set, and obtain the parameter sensitivity matrix; S103: Based on the calibration error and parameter sensitivity matrix, calculate the candidate update amounts for the parameters of the digital twin model within the allowed parameter set; S104: Determine whether the candidate update quantity meets the requirements of maximum single update quantity, minimum update interval, upper limit of cumulative change, and parameter value range based on the parameter update constraints; S105: When the candidate update quantity satisfies the parameter update constraint, candidate digital twin model parameters are generated; when the candidate update quantity does not satisfy the parameter update constraint, the candidate update quantity is not enabled, and the digital twin model parameters before the update remain unchanged. When the preset calibration item corresponds to the sensor measurement error sub-model, the set of parameters that can be updated includes sensor zero-point parameters, sensitivity parameters, range correction parameters, or measurement noise parameters. When the preset calibration item corresponds to the geometric error sub-model, the set of parameters that can be updated includes coordinate mapping parameters, installation posture parameters, measurement direction parameters, motion axis positioning error parameters, straightness error parameters, or backlash parameters. When the preset calibration item corresponds to the thermal error sub-model, the set of allowed update parameters includes the temperature sensitivity coefficient, thermal hysteresis time, thermal drift mapping coefficient, or thermal boundary correction coefficient. When the preset calibration item corresponds to the dynamic error sub-model, the set of parameters that can be updated includes dynamic stiffness parameters, damping parameters, vibration transmission parameters, rotational speed related response parameters, or acceleration / deceleration response parameters. When the preset calibration item corresponds to the process error sub-model, the set of allowed update parameters includes process load mapping parameters, contact stiffness parameters, power mapping parameters, current mapping parameters, or material removal effect parameters. For digital twin model parameters that do not belong to the set of allowed update parameters, their candidate update values ​​are not calculated and the parameter values ​​are kept unchanged.

8. The intelligent calibration and parameter update method for digital twin models of manufacturing equipment as described in claim 1, characterized in that, Steps S11 and S12 include: S111: After obtaining the parameters of the candidate digital twin model, a verification item is selected. The verification item is different from the preset calibration item used for parameter updating; or, the verification item and the preset calibration item used for parameter updating belong to the same type of preset calibration item, but their measurement positions, motion trajectories, running speeds, temperature ranges, load levels, process parameters, or running time periods are different. S112: Perform the verification item and obtain the reference measurement value corresponding to the verification item; S113: Calculate the updated verification error and the original verification error based on the candidate digital twin model parameters and the original digital twin model parameters, respectively; S114: Determine the verification benefit index corresponding to the candidate digital twin model parameters based on the updated verification error, the original verification error, and the candidate update amount within the allowed update parameter set; S121: Determine whether the verification benefit index has reached the preset verification benefit threshold; S122: Based on the parameter update constraints, determine whether the parameters of the candidate digital twin model meet the requirements of maximum single update amount, minimum update interval, upper limit of cumulative change amount, or parameter value range. S123: Based on historical verification results, determine whether the same model parameters, the same error sub-model, or the same preset calibration item have reached the preset historical rollback limit; S124: When the verification benefit index reaches the preset verification benefit threshold, the candidate digital twin model parameters meet the parameter update constraints, and the preset historical rollback limit is not reached, the parameter processing method is determined to be save and enable, the candidate digital twin model parameters are saved and enabled, and a new digital twin model version is generated. S125: When the verification benefit index does not reach the preset verification benefit threshold, or the candidate digital twin model parameters do not meet the parameter update constraints, or the preset historical rollback limit is reached, the parameter processing method is determined to be rollback, rollback to the digital twin model parameters before the update, and the rollback reason is recorded. S126: Update the calibration priority of the preset calibration item according to the parameter processing method, wherein when the parameter processing method is save enabled, increase the calibration priority of the corresponding preset calibration item or record a historical calibration success count; when the parameter processing method is rollback, decrease the calibration priority of the corresponding preset calibration item or record a historical rollback count. S127: When the number of historical rollbacks for the same model parameter, the same error sub-model, or the same preset calibration item reaches the preset number, pause the automatic parameter update of the corresponding model parameter, error sub-model, or preset calibration item, and output a prompt for recalibration or manual review.

Images