Tool wear amount prediction model training method, device and equipment

Abstract

The application discloses a tool wear prediction model training method, device and equipment, comprising: obtaining a tool parameter training sample set corresponding to a first tool wear state of a first tool; inputting an actual tool process parameter into a preset tool wear prediction model to obtain a predicted tool wear; calculating a regression evaluation index according to the predicted tool wear and the actual tool wear; determining whether the regression evaluation index meets a preset training stop condition; if not, adjusting model parameters of the preset tool wear prediction model, and training the adjusted preset tool wear prediction model using the tool parameter training sample set until the preset training stop condition is met, to obtain a trained target tool wear prediction model. Based on this, by continuously adjusting the model parameters and retraining until the preset training stop condition is met, it can be ensured that the target tool wear prediction model finally obtained has high prediction accuracy.

Description

Technical Field

[0001] This application belongs to the field of tool wear prediction, and in particular relates to a training method, apparatus and equipment for a tool wear prediction model. Background Technology

[0002] When machining a workpiece using cutting tools, tool wear is a crucial factor affecting workpiece surface quality and a key factor in tool life. When tool wear becomes excessive and enters a state of rapid wear, cutting force and temperature increase sharply, the contact between the tool face and the workpiece becomes unstable, workpiece surface quality deteriorates, machine tool chatter occurs, and tool failure occurs. By properly setting various parameters in milling operations to keep the tool wear state within the initial wear stage and normal wear stage, a series of problems caused by rapid tool wear can be avoided. This ensures workpiece surface quality, extends tool life, improves production efficiency, and saves production costs.

[0003] Tool wear conditions are typically judged by on-site process engineers based on their personal experience, and tool wear does not change linearly. Therefore, misjudgments of tool wear conditions by on-site process engineers are inevitable. Moreover, the scarcity and high production costs of some alloy materials used in tool manufacturing make judging tool wear conditions solely based on the personal experience of on-site process engineers inefficient and costly.

[0004] Currently, common industry methods for predicting tool wear only aim to anticipate whether the tool will undergo drastic changes to avoid impacting the workpiece. These methods often only predict the trend of tool wear. However, patent CN113408195B proposes a method for predicting tool wear status during the machining process, a multi-step tool wear value prediction method based on tool wear data. Through basic tool wear tests, tool wear curves and data are obtained. The obtained tool wear data is divided into a training set T1 and a test set T2 using a sliding window operation. A dense residual neural network is built and trained to accurately predict tool wear trends. This method can determine whether the tool wear value will change significantly in the future, ensuring that the dimensional accuracy of the machined parts meets usage requirements. Based on the predicted tool wear status during the machining process, a decision is made on whether to replace the tool. This improves production efficiency and reduces costs while ensuring machining quality, which is of great significance for current data-driven intelligent manufacturing.

[0005] The aforementioned patent achieves accurate prediction of tool wear trends by building a dense residual neural network and training it. However, this can only predict the trend of tool wear and cannot determine the specific amount of tool wear. Summary of the Invention

[0006] The embodiment of the present application provides a tool wear amount prediction model training method, device and equipment, and improves the accuracy of tool wear amount prediction.

[0007] According to a first aspect of the present application, the embodiment of the present application provides a tool wear amount prediction model training method, which can include:

[0008] Obtaining a tool parameter training sample set corresponding to a first tool wear state of a first tool, the tool parameter training sample set including a plurality of tool parameter training samples, each tool parameter training sample including an actual tool process parameter and a corresponding actual tool wear amount;

[0009] Inputting the actual tool process parameter into a preset tool wear amount prediction model to obtain a predicted tool wear amount, the preset tool wear amount prediction model corresponding to the first tool wear state;

[0010] Calculating a regression evaluation index according to the predicted tool wear amount and the actual tool wear amount;

[0011] Determining whether the regression evaluation index meets a preset training stop condition;

[0012] If not, adjusting model parameters of the preset tool wear amount prediction model, and training the adjusted preset tool wear amount prediction model using the tool parameter training sample set until the preset training stop condition is met, to obtain a trained target tool wear amount prediction model.

[0013] Optionally, calculating the regression evaluation index according to the predicted tool wear amount and the actual tool wear amount includes:

[0014] Obtaining a training sample quantity of the tool parameter training samples in the tool parameter training sample set;

[0015] Determining a difference square of the predicted tool wear amount and the actual tool wear amount;

[0016] Determining a quotient of the difference square sum and the training sample quantity as a difference square mean value;

[0017] Taking a square root of the difference square mean value to obtain the regression evaluation index.

[0018] Optionally, obtaining the tool parameter training sample set corresponding to the first tool wear state of the first tool further includes:

[0019] Obtaining a plurality of tool parameter training samples of the first tool;

[0020] According to the actual tool wear amount in each tool parameter training sample, determining a tool wear state corresponding to each tool parameter training sample, the tool wear state including the first tool wear state;

[0021] According to the tool wear state corresponding to each tool parameter training sample, the plurality of tool parameter training samples are classified to obtain a tool parameter training sample set corresponding to each tool wear state.

[0022] Optionally, before the plurality of tool parameter training samples of the first tool are obtained, the method comprises:

[0023] Obtaining an initial tool parameter training sample set of the first tool, the initial tool parameter training sample set comprising a plurality of initial tool parameter training samples arranged in time sequence;

[0024] The plurality of initial tool parameter training samples are denoised to obtain the plurality of tool parameter training samples.

[0025] Optionally, the plurality of initial tool parameter training samples are denoised to obtain the plurality of tool parameter training samples, comprising:

[0026] According to the initial tool wear amount in each initial tool parameter training sample, a wear amount difference between a first initial tool wear amount of a first initial tool parameter training sample and a second initial tool wear amount of a second initial tool parameter training sample is calculated, the first initial tool parameter training sample and the second initial tool parameter training sample being adjacent and the first initial tool parameter training sample being arranged after the second initial tool parameter training sample;

[0027] In the case that the wear amount difference is not greater than 0, the first initial tool parameter training sample is determined as a noise tool parameter training sample;

[0028] The noise tool parameter training sample in the initial tool parameter training sample set is deleted to obtain the tool parameter training sample.

[0029] Optionally, the plurality of initial tool parameter training samples are denoised to obtain the plurality of tool parameter training samples, comprising:

[0030] Obtaining a comparative tool parameter sample of a second tool, the second tool being of the same type as the first tool but having a larger diameter than the first tool;

[0031] In the case that the tool wear state corresponding to the comparative tool parameter sample is the same as the tool wear state corresponding to the third initial tool parameter training sample but the comparative tool wear amount in the comparative tool parameter sample is less than the third initial tool wear amount in the third initial tool parameter training sample, the third initial tool parameter training sample is determined as a noise tool parameter training sample;

[0032] The noise tool parameter training sample in the initial tool parameter training sample set is deleted to obtain the tool parameter training sample.

[0033] Optionally, after the actual tool wear amount in each tool parameter training sample is determined according to each tool parameter training sample, and the tool wear state corresponding to each tool parameter training sample is determined, the method further comprises:

[0034] According to the tool wear state corresponding to the tool parameter training sample, the target tool parameter training sample corresponding to the target tool wear state is screened out, the tool wear state includes the target tool wear state, and the target tool wear state includes the first tool wear state.

[0035] According to the tool wear state corresponding to each tool parameter training sample, the plurality of tool parameter training samples are classified to obtain a tool parameter training sample set corresponding to each tool wear state, and the tool parameter training sample set includes:

[0036] According to each target tool wear state, the plurality of target tool parameter training samples are classified to obtain a tool parameter training sample set corresponding to each target tool wear state.

[0037] Optionally, the tool wear state includes an initial wear state, a normal wear state, an acute wear state, and a broken state.

[0038] The target tool wear state includes an initial wear state, a normal wear state, and an acute wear state.

[0039] According to a second aspect of the present application, a training device of a tool wear amount prediction model is provided, which can include:

[0040] The first acquisition module is configured to acquire a tool parameter training sample set corresponding to a first tool wear state of a first tool, the tool parameter training sample set including a plurality of tool parameter training samples, and each tool parameter training sample including an actual tool process parameter and a corresponding actual tool wear amount.

[0041] The input module is configured to input the actual tool process parameter into a preset tool wear amount prediction model to obtain a predicted tool wear amount, the preset tool wear amount prediction model corresponding to the first tool wear state.

[0042] The calculation module is configured to calculate a regression evaluation index according to the predicted tool wear amount and the actual tool wear amount.

[0043] The judgment module is configured to judge whether the regression evaluation index meets a preset training stop condition.

[0044] The loop module is configured to adjust a model parameter of the preset tool wear amount prediction model if the preset training stop condition is not met, and train the adjusted preset tool wear amount prediction model using the tool parameter training sample set until the preset training stop condition is met, to obtain a trained target tool wear amount prediction model.

[0045] According to a third aspect of the present application, the embodiments of the present application provide a training device of a tool wear amount prediction model, comprising a processor and a memory storing computer program instructions.

[0046] The processor implements the tool wear amount prediction model training method according to any one of the first aspect when executing the computer program instructions.

[0047] The embodiments of the present application provide at least the following beneficial effects:

[0048] The embodiments of the present application provide a tool wear amount prediction model training method, device and equipment. A tool parameter training sample set corresponding to a first tool wear state of a first tool is obtained. Actual tool process parameters are input into a preset tool wear amount prediction model to obtain a predicted tool wear amount. A regression evaluation index is calculated according to the predicted tool wear amount and the actual tool wear amount. It is determined whether the regression evaluation index meets a preset training stop condition. If not, the model parameters of the preset tool wear amount prediction model are adjusted, and the preset tool wear amount prediction model after adjustment is trained using the tool parameter training sample set until the preset training stop condition is met, and a target tool wear amount prediction model after training is obtained. Based on this, by using the tool parameter training sample set containing the actual tool process parameters and the corresponding actual tool wear amount, the model can learn the complex relationship between tool wear and process parameters. By continuously adjusting the model parameters and retraining until the preset training stop condition is met, it can be ensured that the target tool wear amount prediction model obtained finally has high prediction accuracy.

[0049] It should be understood that the above general description and the following detailed description are only exemplary and explanatory, and cannot limit the present application. BRIEF DESCRIPTION OF DRAWINGS

[0050] The accompanying drawings incorporated in the specification and forming a part thereof illustrate embodiments consistent with the present application and together with the description are used to explain the principles of the present application, and do not limit the present application.

[0051] Figure 1 is a flowchart of a tool wear amount prediction model training method according to an exemplary embodiment;

[0052] Figure 2 is a structural schematic diagram of a tool wear amount prediction model training device according to an exemplary embodiment;

[0053] Figure 3 is a structural block diagram of a tool wear amount prediction model training device according to an exemplary embodiment. DETAILED DESCRIPTION

[0054] In order to enable a more clear understanding of the above-mentioned objects, features and advantages of the present disclosure, the schemes of the present disclosure will be further described below. It should be noted that the embodiments of the present disclosure and the features in the embodiments can be combined with each other without conflict.

[0055] In the following description, a large number of specific details are set forth in order to facilitate a thorough understanding of the present disclosure, but the present disclosure can also be implemented in other manners different from those described herein; obviously, the embodiments described in the specification are only a part of the embodiments of the present disclosure, and not all the embodiments.

[0056] As described in the background, the tool wear state is usually judged by the on-site process personnel through personal experience, and the tool wear amount does not change linearly, so it is inevitable that the on-site process personnel will misjudge the tool wear state. Moreover, the scarcity and high production cost of some alloy materials for manufacturing tools make it inefficient and costly to judge the tool wear state only by the personal experience of on-site process personnel.

[0057] Based on this, the present application provides a tool wear amount prediction model training method, device and equipment. First, the tool wear amount prediction model training method provided by the embodiments of the present application will be introduced.

[0058] Embodiment 1;

[0059] Figure 1 A flowchart of a tool wear amount prediction model training method provided by an embodiment of the present application is shown. As shown in Figure 1 It can include the following steps:

[0060] S101, obtaining a tool parameter training sample set corresponding to a first tool wear state of a first tool, the tool parameter training sample set including a plurality of tool parameter training samples, the tool parameter training sample including an actual tool process parameter and a corresponding actual tool wear amount;

[0061] S102, inputting the actual tool process parameter into a preset tool wear amount prediction model to obtain a predicted tool wear amount, the preset tool wear amount prediction model corresponding to the first tool wear state;

[0062] S103, calculating a regression evaluation index according to the predicted tool wear amount and the actual tool wear amount;

[0063] S104, judging whether the regression evaluation index meets a preset training stop condition;

[0064] S105, if not satisfied, adjust the model parameters of the preset tool wear amount prediction model, and train the adjusted preset tool wear amount prediction model using the tool parameter training sample set until the preset training stop condition is satisfied, and obtain the trained target tool wear amount prediction model.

[0065] Based on the above embodiment, by obtaining the tool parameter training sample set of the first tool in the first tool wear state; inputting the actual tool process parameters into the preset tool wear amount prediction model to obtain the predicted tool wear amount; calculating the regression evaluation index according to the predicted tool wear amount and the actual tool wear amount; determining whether the regression evaluation index satisfies the preset training stop condition; if not satisfied, adjusting the model parameters of the preset tool wear amount prediction model, and training the adjusted preset tool wear amount prediction model using the tool parameter training sample set until the preset training stop condition is satisfied, and obtaining the trained target tool wear amount prediction model. Based on this, by using the tool parameter training sample set containing the actual tool process parameters and the corresponding actual tool wear amount, the model can learn the complex relationship between tool wear and process parameters. By continuously adjusting the model parameters and retraining until the preset training stop condition is satisfied, it can be ensured that the final obtained target tool wear amount prediction model has high prediction accuracy.

[0066] In the above S101, during the machining of the workpiece by the tool, the tool will inevitably have a certain wear. During the machining of the workpiece by the tool, the tool parameter training sample of the tool wear generated during the machining process can be obtained by combining experiments and simulation means based on the tool wear observation instrument. Therefore, the tool parameter training sample set of the first tool in the first tool wear state can be obtained from the tool wear observation instrument.

[0067] During the service life of the tool, the tool parameter training sample is collected in real time, so the number of tool parameter training samples is necessarily multiple. The multiple tool parameter training samples collected during the service life of each tool form a tool parameter training sample set.

[0068] In each tool parameter training sample, the actual tool process parameters of the tool machining at a certain time and the actual tool wear amount of the tool itself are included. The actual process parameters can include spindle speed, feed per tooth, tool cutting depth, tool cutting width, cutting time, etc.

[0069] In the above S102, the tool process parameters and the tool wear amount are linearly related. Based on this, a preset tool wear amount prediction model can be built by software and programming languages (which can include C, C++, Python). The existing preset tool wear amount prediction model related to tool process parameters and tool wear amount can be directly obtained.

[0070] The collected actual tool process parameters are input into the preset tool wear amount prediction model, and a predicted tool wear amount of the first tool at the actual tool process parameters is predicted through calculation of the preset tool wear amount prediction model.

[0071] In an example, the software can include any one of MATLAB, Visual Studio, Xcode, etc., and the programming language can include any one of C, C++, Python, etc.

[0072] It should be noted that the linear relationship between the tool process parameters and the tool wear amount of the tool is different in different wear states, and therefore, the corresponding tool wear amount prediction model is also different in different wear states. The tool wear amount prediction model selected for the first tool in the first tool wear state is the preset tool wear amount prediction model corresponding to the first tool wear state.

[0073] In the above S103, after obtaining the predicted tool wear amount predicted by the preset tool wear amount prediction model, the accuracy of the predicted tool wear amount can be determined by calculating the regression evaluation index between the predicted tool wear amount and the actual tool wear amount.

[0074] Specifically, the above S103 can include:

[0075] S1031, obtaining a training sample number of tool parameter training samples in a tool parameter training sample set;

[0076] S1032, determining a difference square of the predicted tool wear amount and the actual tool wear amount;

[0077] S1033, determining a quotient of the difference square and the training sample number as a difference square mean value;

[0078] S1034, taking a square root of the difference square mean value to obtain the regression evaluation index.

[0079] In the above S1031, the number of tool parameter training samples in the tool parameter training set is counted to obtain the training sample number.

[0080] In the above S1032, a prediction difference value between the predicted tool wear amount and the actual tool wear amount is calculated, and then a square root of the prediction difference value is calculated to obtain the difference square.

[0081] In the above S1033, the difference square is divided by the training sample number to obtain the difference square mean value.

[0082] In the above S1034, the square root of the difference square mean value is calculated, and the obtained value is the regression evaluation index.

[0083] More specifically, the steps of S1032-S1034 above can be implemented by the following formula.

[0084]

[0085] wherein y i is the predicted tool wear amount, is the actual tool wear amount, and m is the number of training samples.

[0086] In the above S104, it is determined whether the regression evaluation index meets a preset training stop condition, wherein the preset training stop condition refers to that the regression evaluation index is less than a preset threshold, and in the present embodiment, the preset threshold can be 10 μm.

[0087] If yes, it can be considered that the preset tool wear amount prediction model is accurate, and is the target tool wear amount prediction model that has been trained.

[0088] In the above S105, if no, the model parameters in the preset tool wear amount prediction model are adjusted, and the tool parameter training samples in the tool parameter training sample set are input into the adjusted preset tool wear amount prediction model again, the steps of S102-S103 above are repeated for training until the new regression evaluation index obtained can meet the preset training stop condition, so that the preset tool wear amount prediction model finally obtained has been trained, and the trained preset tool wear amount prediction model is the target tool wear amount prediction model.

[0089] In order to accurately distinguish the wear state of the tool, the present application further provides another implementation manner of the tool wear amount prediction model training method.

[0090] Before the above S101, the tool wear amount prediction model training method can further include:

[0091] S201, obtaining a plurality of tool parameter training samples of a first tool;

[0092] S202, determining the tool wear state corresponding to each tool parameter training sample according to the actual tool wear amount in each tool parameter training sample, wherein the tool wear state includes a first tool wear state.

[0093] S203, classifying the plurality of tool parameter training samples according to the tool wear state corresponding to each tool parameter training sample, to obtain a tool parameter training sample set corresponding to each tool wear state.

[0094] Based on the above embodiment, by acquiring a plurality of tool parameter training samples of the first tool in the wear test process, the actual tool wear amount in each tool parameter training sample can be analyzed, and the tool wear state of the first tool at each tool wear amount can be determined according to the tool wear amount. Therefore, the plurality of tool parameter training samples can be classified according to the tool wear state, so that each different tool wear state has a corresponding tool parameter training sample set. Based on this, by classifying the different tool training parameter samples obtained according to the tool wear amount, different tool parameter training sample sets are formed, so that each different tool parameter training sample set can be used to train different preset tool wear amount prediction models in the future, and the accuracy of training different preset tool wear amount prediction models is improved.

[0095] In the above S201, all tool parameter training samples collected by the tool wear observation instrument during the entire machining process of the first tool are acquired.

[0096] Optionally, in an embodiment, the above S201 can further include:

[0097] S2011, acquiring an initial tool parameter training sample set of the first tool, the initial tool parameter training sample set including a plurality of initial tool parameter training samples arranged in time sequence;

[0098] S2012, denoising the plurality of initial tool parameter training samples to obtain a plurality of tool parameter training samples.

[0099] Based on the above embodiment, by denoising the acquired initial tool parameter training samples, the noise data in the initial tool parameter training samples is removed, so that the plurality of tool parameter training samples obtained finally are more accurate, and the accuracy of training the preset tool wear amount prediction model is improved.

[0100] In the above S2011, the initial tool parameter training sample set directly acquired from the tool wear observation instrument can be included in the plurality of initial tool parameter training samples. The initial tool parameter training sample refers to the initial tool parameter training sample directly acquired from the tool wear observation instrument without various processing.

[0101] It should be noted that since the initial tool parameter training sample of the tool wear observation instrument recording the wear of the tool in the machining process is also real-time, each initial tool parameter training sample in the initial tool parameter training sample set is arranged in time sequence.

[0102] In the S2012, the tool wear observer records the tool wear state, and the environment and the error of the tool wear observer itself exist, so that the initial tool parameter training sample set contains errors. Therefore, the initial tool parameter training sample in the initial tool parameter training sample set is denoised to remove the errors, and the initial tool parameter training sample left in the initial tool parameter training sample set is the tool parameter training sample after denoising.

[0103] Specifically, in an embodiment, the S2012 can include:

[0104] S20121, calculating a wear amount difference between a first initial tool wear amount of a first initial tool parameter training sample and a second initial tool wear amount of a second initial tool parameter training sample, the first initial tool parameter training sample and the second initial tool parameter training sample being adjacent and the first initial tool parameter training sample being arranged after the second initial tool parameter training sample according to the initial tool wear amount in each initial tool parameter training sample;

[0105] S20122, determining the first initial tool parameter training sample as a noise tool parameter training sample in a case that the wear amount difference is not greater than 0;

[0106] S20123, deleting the noise tool parameter training sample in the initial tool parameter training sample set to obtain the tool parameter training sample.

[0107] In the S20121, the initial tool parameter training samples in the initial tool parameter training sample set are arranged in time sequence, so that the wear amount difference is obtained by calculating the difference between the first initial tool wear amount and the second initial tool wear amount of the first tool, and the wear amount difference can reflect the wear of the tool in the process of machining the workpiece.

[0108] In the S20122, the tool will be worn in the process of machining the workpiece, so the wear amount difference is greater than 0. Therefore, if the wear amount difference is not greater than 0, it is contrary to the objective fact rule, so the first initial tool parameter training sample is a noise tool parameter training sample.

[0109] In the S20123, the noise tool parameter training sample in the initial tool parameter training sample set is deleted, so that a more accurate tool parameter training sample is obtained.

[0110] More specifically, in another embodiment, the S2012 can include:

[0111] S20124, obtaining a contrast tool parameter sample of a second tool, the second tool being of the same type as the first tool but having a larger diameter than the first tool;

[0112] S20125, determining that the third initial tool parameter training sample is a noise tool parameter training sample, in a case that the contrast tool process parameter of the contrast tool parameter sample is the same as the actual tool process parameter of the third initial tool parameter training sample, but the contrast tool wear amount in the contrast tool parameter sample is less than the third initial tool wear amount in the third initial tool parameter training sample;

[0113] S20126, deleting the noise tool parameter training sample in the initial tool parameter training sample set to obtain the tool parameter training sample.

[0114] In the above S20124, the contrast tool parameter sample of the second tool of the same tool type can also be obtained from the same tool wear observer that obtains the initial tool parameter training sample of the first tool. It should be noted that the tool diameter of the second tool is larger than that of the first tool.

[0115] The tool type can include turning tools, planing tools, milling tools, boring tools, reaming tools, etc.

[0116] In the above S20125, the third initial tool training sample with the same contrast tool process parameter as the contrast tool parameter sample is selected from the initial tool sample set. The actual tool process parameter of the third initial tool training sample is the same as the contrast tool process parameter. The sizes of the contrast tool wear amount and the third initial tool wear amount are compared. If the contrast tool wear amount is smaller than the third initial tool wear amount, it does not conform to the objective fact rule, and the third initial tool wear amount has obvious problems. Therefore, the third initial tool parameter training sample corresponding to the third initial tool wear amount is a noise tool parameter training sample.

[0117] In the above S2016, the noise tool parameter training sample of the initial tool parameter training sample set is deleted, thereby obtaining a more accurate tool parameter training sample.

[0118] In the above S202, the tool parameter training sample includes the actual tool wear amount of the first tool. Each tool wear state corresponds to a tool wear amount interval. The tool wear state corresponding to each tool parameter training sample is determined by determining the tool wear amount interval to which each actual tool wear amount belongs according to the size of the actual tool wear amount in each tool parameter training sample.

[0119] As an example, the tool wear amount interval corresponding to the initial wear state is between 0mm and 0.1mm; the tool wear amount interval corresponding to the normal wear state is between 0.1mm and 0.2mm; the tool wear amount interval corresponding to the sharp wear state is between 0.2mm and 0.4mm; and if the tool wear amount is greater than 0.4mm, the tool is in a broken state.

[0120] In S203, each tool parameter training sample is classified according to the tool wear state corresponding to each tool parameter training sample, and tool parameter training samples of the same tool wear state are combined to form a tool parameter training sample set.

[0121] To accurately distinguish between different tool wear states, the application further provides another implementation of the tool wear amount prediction model training method.

[0122] After S202, the tool wear amount prediction model training method can further include:

[0123] S301, target tool parameter training samples corresponding to a target tool wear state are selected according to the tool wear state corresponding to the tool parameter training samples, the tool wear state includes the target tool wear state, and the target tool wear state includes a first tool wear state;

[0124] S202 can include:

[0125] S302, classify the plurality of target tool parameter training samples according to each target tool wear state to obtain a tool parameter training sample set corresponding to each target tool wear state.

[0126] In S301, after determining the tool wear state of each tool parameter training sample in the tool parameter training sample set, target tool parameter training samples with the target wear state are selected from the plurality of tool parameter training samples according to the tool wear state.

[0127] The tool wear state includes an initial wear state, a normal wear state, a sharp wear state, and a broken state.

[0128] The target tool wear state includes an initial wear state, a normal wear state, and a sharp wear state.

[0129] Since the broken state has no value for the tool, the tool parameter training sample corresponding to the broken state can be used to train the preset tool wear amount prediction model, so target tool parameter training samples with the target wear state are selected from the plurality of tool parameter training samples according to the tool wear state.

[0130] If the predicted tool wear amount obtained by predicting according to the tool process parameters is not in the tool wear amount interval corresponding to the initial wear state, the normal wear state, or the sharp wear state, the predicted tool wear amount can be directly output as the broken state without outputting the predicted tool wear amount.

[0131] In the above S302, the plurality of target tool parameter training samples in the target tool wear state are classified according to the types of the target tool wear state, so that a plurality of tool parameter training samples corresponding to each target tool wear state are classified, and each target tool wear state corresponds to a tool parameter training sample set.

[0132] As an example, after the target tool wear amount prediction model corresponding to each target wear state is trained by the above method, the tool process parameters can be used as the input parameters of the target tool wear amount prediction model corresponding to the initial wear state to calculate the corresponding predicted tool wear amount. To ensure that the predicted tool wear amount is predicted to be too small, the tool process parameters are further input into the target tool wear amount prediction model corresponding to the normal wear state for secondary verification. If the predicted tool wear amount calculated by the target tool wear amount prediction model corresponding to the normal wear state is obviously less than 0.11 mm, it indicates that the target tool wear amount prediction model corresponding to the initial wear state is correct, and the predicted tool wear amount can be used as the actual tool wear amount under the tool process parameters.

[0133] If the predicted tool wear amount calculated by the target tool wear amount prediction model corresponding to the normal wear state is obviously greater than 0.11 mm, it indicates that the tool wear state classification is incorrect or the target tool wear amount prediction model corresponding to the initial wear state is incorrect. At this time, the predicted tool wear amount calculated by the target tool wear amount prediction model corresponding to the normal wear state is used as the reference. To ensure the accuracy of the predicted tool wear amount calculated by the target tool wear amount prediction model corresponding to the normal wear state, the tool process parameters are further input into the target tool wear amount prediction model corresponding to the sharp wear state for secondary verification. If the predicted tool wear amount calculated by the target tool wear amount prediction model corresponding to the sharp wear state is obviously less than 0.21 mm, it indicates that the predicted tool wear amount calculated by the target tool wear amount prediction model corresponding to the normal wear state is correct, and the predicted tool wear amount calculated by the target tool wear amount prediction model corresponding to the normal wear state can be used as the tool predicted tool wear amount under the tool process parameters.

[0134] If the predicted tool wear amount calculated by the target tool wear amount prediction model corresponding to the sharp wear state is obviously greater than 0.21 m, it indicates that the target tool wear amount prediction model corresponding to the normal wear state has prediction error. At this time, the predicted tool wear amount calculated by the target tool wear amount prediction model corresponding to the sharp wear state is used as the criterion. If the predicted tool wear amount is greater than 0.4 mm, it is directly output as a broken state.

[0135] Embodiment 2;

[0136] Based on the same inventive concept, the embodiment further provides a tool wear amount prediction model training device 200 based on the embodiment 1, which is specifically as shown in Figure 2 The tool wear amount prediction model training device 200 comprises:

[0137] The first acquisition module 210 is configured to acquire a tool parameter training sample set corresponding to a first tool wear state of a first tool, wherein the tool parameter training sample set comprises a plurality of tool parameter training samples, and each tool parameter training sample comprises an actual tool process parameter and a corresponding actual tool wear amount.

[0138] The input module 220 is configured to input the actual tool process parameter into a preset tool wear amount prediction model to obtain a predicted tool wear amount, wherein the preset tool wear amount prediction model corresponds to the first tool wear state.

[0139] The calculation module 230 is configured to calculate a regression evaluation index according to the predicted tool wear amount and the actual tool wear amount.

[0140] The judgment module 240 is configured to judge whether the regression evaluation index meets a preset training stop condition.

[0141] The loop module 250 is configured to adjust model parameters of the preset tool wear amount prediction model if the regression evaluation index does not meet the preset training stop condition, and train the adjusted preset tool wear amount prediction model by using the tool parameter training sample set until the preset training stop condition is met, so as to obtain a trained target tool wear amount prediction model.

[0142] Based on the above embodiment, the first cutter parameter training sample set corresponding to the first cutter wear state of the first cutter is obtained by the first acquisition module 210; the actual cutter process parameter is input into the preset cutter wear amount prediction model by the input module 220, and the predicted cutter wear amount is obtained; the regression evaluation index is calculated according to the predicted cutter wear amount and the actual cutter wear amount by the calculation module 230; whether the regression evaluation index meets the preset training stop condition is judged by the judgment module 240; if not, the model parameters of the preset cutter wear amount prediction model are adjusted, and the preset cutter wear amount prediction model after adjustment is trained by using the cutter parameter training sample set, until the preset training stop condition is met, and the trained target cutter wear amount prediction model is obtained. Based on this, by using the cutter parameter training sample set containing the actual cutter process parameter and the corresponding actual cutter wear amount, the model can learn the complex relationship between the cutter wear and the process parameter. By continuously adjusting the model parameters and retraining until the preset training stop condition is met, it can be ensured that the target cutter wear amount prediction model obtained finally has high prediction accuracy.

[0143] Optionally, the calculation module 230 can include:

[0144] The first acquisition unit is configured to acquire the training sample quantity of the cutter parameter training sample in the cutter parameter training sample set.

[0145] The first determination unit is configured to determine the difference square of the predicted cutter wear amount and the actual cutter wear amount.

[0146] The second determination unit is configured to determine the quotient of the difference square and the training sample quantity as the difference square mean.

[0147] The square root unit is configured to take the square root of the difference square mean to obtain the regression evaluation index.

[0148] Optionally, the first acquisition module 210 can further include:

[0149] The second acquisition unit is configured to acquire a plurality of cutter parameter training samples of the first cutter.

[0150] The third determination unit is configured to determine the cutter wear state corresponding to each cutter parameter training sample according to the actual cutter wear amount in each cutter parameter training sample, and the cutter wear state includes the first cutter wear state.

[0151] The classification unit is configured to classify the plurality of cutter parameter training samples according to the cutter wear state corresponding to each cutter parameter training sample to obtain the cutter parameter training sample set corresponding to each cutter wear state.

[0152] Optionally, the training device 200 of the cutter wear amount prediction model can further include:

[0153] The second acquisition module is configured to acquire an initial tool parameter training sample set of the first tool, the initial tool parameter training sample set comprising a plurality of initial tool parameter training samples arranged in a time sequence.

[0154] The denoising module is configured to perform denoising processing on the plurality of initial tool parameter training samples to obtain a plurality of tool parameter training samples.

[0155] Optionally, the denoising module can comprise:

[0156] The calculation unit is configured to calculate, according to the initial tool wear amount in each initial tool parameter training sample, a wear amount difference between a first initial tool wear amount of a first initial tool parameter training sample and a second initial tool wear amount of a second initial tool parameter training sample, the first initial tool parameter training sample and the second initial tool parameter training sample being adjacent and the first initial tool parameter training sample being arranged after the second initial tool parameter training sample.

[0157] The fourth determination unit is configured to determine that the first initial tool parameter training sample is a noise tool parameter training sample in a case where the wear amount difference is not greater than 0.

[0158] The first deletion module is configured to delete the noise tool parameter training sample in the initial tool parameter training sample set to obtain the tool parameter training sample.

[0159] Optionally, the denoising module can further comprise:

[0160] The third acquisition unit is configured to acquire a comparison tool parameter sample of a second tool, the second tool being of the same type as the first tool but having a larger diameter than the first tool.

[0161] The fifth determination unit is configured to determine that the third initial tool parameter training sample is a noise tool parameter training sample in a case where the tool wear state corresponding to the comparison tool parameter sample is the same as the tool wear state corresponding to the third initial tool parameter training sample, but the comparison tool wear amount in the comparison tool parameter sample is less than the third initial tool wear amount in the third initial tool parameter training sample.

[0162] The second deletion unit is configured to delete the noise tool parameter training sample in the initial tool parameter training sample set to obtain the tool parameter training sample.

[0163] Optionally, the first acquisition module 210 can further comprise:

[0164] The screening unit is configured to screen target tool parameter training samples corresponding to a target tool wear state according to tool parameter training samples corresponding to tool wear states, wherein the tool wear states include the target tool wear state, and the target tool wear state includes a first tool wear state.

[0165] The classification unit can include:

[0166] The classification sub-unit is configured to classify the plurality of target tool parameter training samples according to each target tool wear state, to obtain a tool parameter training sample set corresponding to each target tool wear state.

[0167] Optionally, the tool wear states include an initial wear state, a normal wear state, an acute wear state and a broken state.

[0168] The target tool wear state includes an initial wear state, a normal wear state and an acute wear state.

[0169] In the embodiment 2, the processes of the above-mentioned embodiment 1 can be implemented, and the same technical effects can be achieved. To avoid repetition, details are not described herein.

[0170] Embodiment 3:

[0171] Figure 3 A hardware structure schematic diagram of a training device of a tool wear amount prediction model is shown.

[0172] The training device of the tool wear amount prediction model can include a processor 301 and a memory 302 storing computer program instructions.

[0173] Specifically, the processor 301 can include a central processing unit (CPU), or an application specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of the present application.

[0174] The memory 302 can include a mass storage for data or instructions. By way of example and not limitation, the memory 302 can include a hard disk drive (HDD), a floppy disk drive, a flash memory, an optical disk, a magneto-optical disk, a magnetic tape, or a universal serial bus (USB) drive or a combination of two or more of these. Where appropriate, the memory 302 can include removable or non-removable (or fixed) media. Where appropriate, the memory 302 can be internal or external to the integrated gateway disaster recovery device. In a particular embodiment, the memory 32 is a non-volatile solid-state memory.

[0175] In particular embodiments, memory 302 can include read-only memory (ROM), random-access memory (RAM), magnetic disk storage mediums, optical storage mediums, flash memory devices, electrical, optical, or other physical / tangible memory storage devices. Thus, generally, memory 302 includes one or more tangible (non-transitory) computer-readable storage media (e.g., a memory device) encoded with software that, when executed (e.g., by one or more processors 301), is operable to perform the operations described with reference to the methods according to an aspect of the present application.

[0176] Processor 301 implements the training method of any of the above-described tool wear amount prediction models by reading and executing computer program instructions stored in memory 302.

[0177] In one example, the tool wear amount prediction model training device can further include a communication interface 303 and a bus 304. Wherein, as shown in the figure, the processor 301, the memory 302, the communication interface 303 are connected through the bus 304 and complete the communication between each other.

[0178] The communication interface 303 is mainly used to realize the communication between the modules, devices, units and / or devices in the embodiments of the application.

[0179] The bus 304 includes hardware, software, or both. By way of example and not limitation, the bus 304 can include an accelerated graphics port (AGP) or other graphics bus, an enhanced industry standard architecture (EISA) bus, a front-side bus (FSB), a hyper transport (HT) interconnect, an industry standard architecture (ISA) bus, an inter-processor interrupt (IPI) bus, a low pin count (LPC) bus, a memory bus, a micro channel architecture (MCA) bus, a peripheral component interconnect (PCI) bus, a PCI-Express (PCI-X) bus, a serial advanced technology attachment (SATA) bus, a video electronics standards association local (VLB) bus, or other suitable bus or combination of two or more of these. Where appropriate, the bus 304 can include one or more buses 304. Although particular buses 304 are described and illustrated in this application, the application contemplates any suitable bus 304 or interconnect.

[0180] The tool wear amount prediction model training device can be based on the current tool wear amount prediction model training method, so as to realize the tool wear amount prediction model training method and the tool wear amount prediction model training device 200 described in combination Figure 1 、 2 The tool wear amount prediction model training method and the tool wear amount prediction model training device 200 described in combination

[0181] In addition, an embodiment of the present application further provides a computer program product comprising computer program instructions, and the computer program product, when executed by the processor 301, can implement the steps and corresponding contents of the foregoing method embodiments.

[0182] It should be noted that the present application is not limited to the specific configurations and processes described above and shown in the drawings. For the sake of brevity, detailed descriptions of well-known methods are omitted. In the above embodiments, several specific steps are described and shown as examples. However, the method processes of the present application are not limited to the specific steps described and shown, and those skilled in the art can make various changes, modifications and additions, or change the order of the steps, after understanding the spirit of the present application.

[0183] The functional blocks shown in the structural block diagrams described above can be implemented as hardware, software, firmware or a combination thereof. When implemented in hardware, it can be, for example, an electronic circuit, an application specific integrated circuit (ASIC), appropriate firmware, a plug-in, a functional card, etc. When implemented in software, the elements of the present application are program or code segments used to perform the required tasks. The program or code segments can be stored in a machine-readable medium or transmitted through a data signal carried in a carrier wave over a transmission medium or communication link. The "machine-readable medium" can include any medium capable of storing or transmitting information. Examples of machine-readable media include electronic circuits, semiconductor memory devices, ROM, flash memory, erasable ROM (EROM), floppy disks, CD-ROMs, optical disks, hard disks, optical fiber media, radio frequency (RF) links, etc. The code segments can be downloaded via a computer network such as the Internet, an intranet, etc.

[0184] It should also be noted that the exemplary embodiments mentioned in the present application describe some methods or systems based on a series of steps or devices. However, the present application is not limited to the order of the above steps, that is, the steps can be executed in the order mentioned in the embodiments, or in an order different from the embodiments, or several steps can be executed simultaneously.

[0185] The computer program instructions can also be loaded onto a computer, other programmable data processing apparatus, or other programmable data processing devices to cause a series of operational steps to be performed on the computer, other programmable apparatus or other programmable devices to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide processes for implementing the functions / acts specified in the flowchart and / or block diagram block or blocks.

[0186] The above describes only specific implementation of the present application. For the convenience and brevity of description, the specific working process of the system, module and unit described above can refer to the corresponding process in the foregoing method embodiments, which will not be described herein. It should be understood that the protection scope of the present application is not limited in this way. Any person skilled in the art can easily think of various equivalent modifications or replacements within the technical range disclosed in the present application, and these modifications or replacements should be covered within the protection scope of the present application.

Claims

1. A training method for a tool wear prediction model, characterized in that, The method includes: Obtain a tool parameter training sample set corresponding to the first tool wear state. The tool parameter training sample set includes multiple tool parameter training samples, and the tool parameter training samples include actual tool process parameters and corresponding actual tool wear amounts. The actual tool process parameters are input into a preset tool wear prediction model to obtain the predicted tool wear amount. The preset tool wear prediction model corresponds to the first tool wear state. The regression evaluation index is calculated based on the predicted tool wear and the actual tool wear. Determine whether the regression evaluation index meets the preset training stopping condition; If not satisfied, adjust the model parameters of the preset tool wear prediction model, and use the tool parameter training sample set to train the adjusted preset tool wear prediction model until the preset training stop condition is met, and obtain the trained target tool wear prediction model. Before obtaining the tool parameter training sample set corresponding to the first tool wear state, the method further includes: Obtain multiple training samples of the tool parameters of the first tool; Based on the actual tool wear in each tool parameter training sample, determine the tool wear state corresponding to each tool parameter training sample, wherein the tool wear state includes the first tool wear state; Based on the tool wear state corresponding to each tool parameter training sample, the multiple tool parameter training samples are classified to obtain a tool parameter training sample set corresponding to each tool wear state; The step of obtaining multiple training samples of the tool parameters of the first tool includes: Obtain an initial tool parameter training sample set for the first tool, wherein the initial tool parameter training sample set includes multiple initial tool parameter training samples arranged in chronological order of acquisition. Denoising is performed on multiple initial tool parameter training samples to obtain multiple tool parameter training samples. The step of denoising multiple initial tool parameter training samples to obtain multiple tool parameter training samples includes: Based on the initial tool wear in each initial tool parameter training sample, calculate the wear difference between the first initial tool wear of the first initial tool parameter training sample and the second initial tool wear of the second initial tool parameter training sample, wherein the first initial tool parameter training sample and the second initial tool parameter training sample are adjacent, and the first initial tool parameter training sample is arranged after the second initial tool parameter training sample. If the wear difference is not greater than 0, the first initial tool parameter training sample is determined to be a noisy tool parameter training sample. The noisy tool parameter training samples in the initial tool parameter training sample set are deleted to obtain the tool parameter training samples. The step of denoising multiple initial tool parameter training samples to obtain multiple tool parameter training samples may include: Obtain a sample of comparative tool parameters for the second tool. The second tool is of the same type as the first tool, but the diameter of the second tool is larger than that of the first tool. If the comparative tool process parameters in the comparative tool parameter sample are the same as the actual tool process parameters in the third initial tool parameter training sample, but the wear amount of the comparative tool in the comparative tool parameter sample is less than the wear amount of the third initial tool in the third initial tool parameter training sample, then the third initial tool parameter training sample is determined to be a noisy tool parameter training sample. The noisy tool parameter training samples in the initial tool parameter training sample set are deleted to obtain the tool parameter training samples.

2. The method as described in claim 1, characterized in that, The regression evaluation index calculated based on the predicted tool wear and the actual tool wear includes: Obtain the number of training samples in the tool parameter training sample set; Determine the squared difference between the predicted tool wear and the actual tool wear; The quotient of the squared difference and the number of training samples is defined as the mean of the squared difference. The regression evaluation index is obtained by taking the square root of the squared mean of the difference.

3. The method as described in claim 1, characterized in that, After determining the tool wear state corresponding to each tool parameter training sample based on the actual tool wear amount in each tool parameter training sample, the method further includes: Based on the tool wear state corresponding to the tool parameter training sample, a target tool parameter training sample corresponding to the target tool wear state is obtained by filtering. The tool wear state includes the target tool wear state, and the target tool wear state includes the first tool wear state. The step of classifying multiple tool parameter training samples according to the tool wear state corresponding to each tool parameter training sample to obtain a tool parameter training sample set corresponding to each tool wear state includes: Based on each of the target tool wear states, multiple target tool parameter training samples are classified to obtain a tool parameter training sample set corresponding to each of the target tool wear states.

4. The method as described in claim 3, characterized in that, The tool wear conditions include: initial wear condition, normal wear condition, rapid wear condition, and broken condition; The wear state of the target tool includes: initial wear state, normal wear state, and rapid wear state.

5. A training device for a tool wear prediction model, characterized in that, The device includes: a processor and a memory storing computer program instructions; When the processor executes the computer program instructions, it implements the training method for the tool wear prediction model as described in any one of claims 1-4.

Images