Machine tool prediction method and apparatus, and terminal device and computer-readable storage medium

Abstract

A machine tool prediction method and apparatus, and a terminal device and a computer-readable storage medium. The method comprises: acquiring operation data of a numerical control machine tool on the basis of a sampling frequency for the numerical control machine tool, wherein the operation data at least comprises temperature data, pressure data and vibration data of the numerical control machine tool at the current moment, and the sampling frequency is used for determining a time interval from the current moment to the next moment (S10); determining a student model specified by the numerical control machine tool, wherein the student model refers to a model that performs training and learning on a trained teacher model according to a preset knowledge distillation algorithm, and the student model and the teacher model are both state detection models of the numerical control machine tool (S20); separately inputting the operation data into the student model and the teacher model, so as to obtain an actual predicted state of the numerical control machine tool that is output by the student model, and a given real state of the numerical control machine tool that is output by the teacher model (S30); and determining a real operation state of the numerical control machine tool according to the actual predicted state and the given real state (S40).

Description

Machine tool prediction method, device, terminal device and computer-readable storage medium

[0001] Related applications

[0002] This application claims priority to Chinese patent application No. 202310878111.1 filed on July 18, 2023, the entire contents of which are incorporated herein by reference. Technical Field

[0003] The present application relates to the field of numerical control technology, and in particular to a machine tool prediction method, apparatus, terminal device, and computer-readable storage medium. Background Art

[0004] CNC technology is an indispensable technology in today's advanced manufacturing industry. As CNC technology develops rapidly, users have put forward higher requirements on the prediction accuracy of CNC machine tools.

[0005] Traditional CNC machine tool monitoring systems primarily use predetermined thresholds or rules to predict machine tool operating data. However, this approach has significant drawbacks. The operating environment and state of CNC machine tools change over time and with usage. Using these thresholds or rules to predict CNC machine tool operating data can easily lead to distortion in the predicted machine tool state output by the CNC machine tool monitoring system. This means that the predicted machine tool state is of no use during the production process. Therefore, improving the accuracy of machine tool operating state prediction is a pressing technical issue.

[0006] Summary of the Invention

[0007] The main purpose of this application is to provide a machine tool prediction method, apparatus, terminal device and computer-readable storage medium, aiming to improve the accuracy of predicting the operating status of a machine tool.

[0008] To achieve the above objectives, the present application provides a machine tool prediction method, which includes:

[0009] Acquiring operating data of the CNC machine tool based on a sampling frequency of the CNC machine tool, the operating data including at least temperature data, pressure data, and vibration data of the CNC machine tool at a current moment, the sampling frequency being used to determine a time interval between the current moment and a next moment;

[0010] Determine a student model specified for the CNC machine tool, where the student model is a model that is trained to learn from a pre-trained teacher model according to a preset knowledge distillation algorithm, and both the student model and the teacher model are state detection models of the CNC machine tool;

[0011] Inputting the operating data into the student model and the teacher model respectively, obtaining an actual predicted state of the CNC machine tool output by the student model and a given real state of the CNC machine tool output by the teacher model;

[0012] The actual operating state of the CNC machine tool is determined according to the actual predicted state and the given real state.

[0013] In one embodiment, the step of determining the actual operating state of the CNC machine tool based on the actual predicted state and the given real state includes:

[0014] When the actual predicted state is the same as the given real state, taking the actual predicted state as the real operating state of the CNC machine tool;

[0015] When the actual predicted state is different from the given real state, prediction error data between the actual predicted state and the given real state is obtained, a matching result between the actual predicted state and the given real state is determined based on the prediction error data, and the real operating state of the CNC machine tool is determined based on the matching result and the actual predicted state.

[0016] In one embodiment, the actual predicted state includes at least a temperature predicted state, a pressure predicted state, and a vibration predicted state of the CNC machine tool, and the given real state includes at least a given temperature state, a given pressure state, and a given vibration state of the CNC machine tool;

[0017] The step of obtaining prediction error data between the actual predicted state and the given true state comprises:

[0018] determining a temperature error value between the predicted temperature state and the given temperature state, a pressure error value between the predicted pressure state and the given pressure state, and a vibration error value between the predicted vibration state and the given vibration state;

[0019] Determining a temperature mean square deviation of the CNC machine tool based on the temperature error value, determining a pressure mean square deviation of the CNC machine tool based on the pressure error value, and determining a vibration mean square deviation of the CNC machine tool based on the vibration error value;

[0020] The temperature mean square error, the pressure mean square error, and the vibration mean square error are sequentially superimposed to obtain prediction error data.

[0021] In one embodiment, the step of determining a matching result between the actual predicted state and the given real state based on the prediction error data includes:

[0022] determining an error threshold range specified for the given real state, and detecting whether the predicted error data exceeds the error threshold range;

[0023] If the prediction error data does not exceed the error threshold range, determining that the matching result between the actual prediction state and the given real state is that the actual prediction state matches the given real state;

[0024] If the prediction error data exceeds the error threshold range, it is determined that the matching result between the actual prediction state and the given real state is that the actual prediction state does not match the given real state.

[0025] In one embodiment, the step of determining the actual operating state of the CNC machine tool based on the matching result and the actual predicted state includes:

[0026] When the matching result is that the actual predicted state matches the given true state, the student model is updated according to the loss function specified by the prediction error data to obtain an updated student model, and the step of inputting the operating data into the student model and the teacher model is returned to be executed respectively according to the updated student model until a target predicted state identical to the given true state is obtained through training, and the target predicted state is used as the true operating state of the CNC machine tool;

[0027] When the matching result is that the actual predicted state does not match the given real state, the sampling frequency is adjusted according to the actual predicted state to obtain the adjusted sampling frequency, and the step of obtaining the operating data of the CNC machine tool based on the sampling frequency of the CNC machine tool is returned to based on the adjusted sampling frequency.

[0028] In one embodiment, the step of updating the student model according to the loss function specified by the prediction error data includes:

[0029] determining gradient data of the CNC machine tool according to a loss function specified by the prediction error data, and determining a learning frequency specified by the student model;

[0030] The student model is updated according to the product data between the gradient data and the learning frequency to obtain an updated student model.

[0031] In one embodiment, the machine tool prediction method further includes:

[0032] Building a digital twin model of the CNC machine tool based on the equipment data of the CNC machine tool, transmitting the operating data to the digital twin model, and outputting the actual operating status of the CNC machine tool, wherein the equipment data includes at least machine status data, machine tool tool wear data, and machine workpiece data;

[0033] After the digital twin model obtains the real operating state, determining whether the actual operating state is consistent with the real operating state;

[0034] If the actual operating state is consistent with the true operating state, it is determined that the CNC machine tool is in a normal operating state;

[0035] If the actual operating state is inconsistent with the true operating state, it is determined that the CNC machine tool is in an abnormal operating state, and a fault prompt is given according to the abnormal operating state.

[0036] In addition, to achieve the above-mentioned purpose, the present application also provides a machine tool prediction device, which includes:

[0037] an acquisition module, configured to acquire operating data of the CNC machine tool based on a sampling frequency of the CNC machine tool, the operating data including at least temperature data, pressure data, and vibration data of the CNC machine tool at a current moment, the sampling frequency being used to determine a time interval between the current moment and a next moment;

[0038] a designation module, configured to determine a student model designated by the CNC machine tool, wherein the student model is a model that is trained to learn from a pre-trained teacher model according to a preset knowledge distillation algorithm, and both the student model and the teacher model are state detection models of the CNC machine tool;

[0039] a model output module, inputting the operating data into the student model and the teacher model respectively, to obtain the actual predicted state of the CNC machine tool output by the student model and the given real state of the CNC machine tool output by the teacher model;

[0040] The prediction module determines the actual operating state of the CNC machine tool according to the actual predicted state and the given real state.

[0041] Each functional module of the machine tool prediction device of the present application implements the steps of the machine tool prediction method of the present application as described above during operation.

[0042] In addition, to achieve the above-mentioned purpose, the present application also provides a terminal device, which includes a memory, a processor, and a machine tool prediction program stored on the memory and runnable on the processor. When the machine tool prediction program is executed by the processor, the steps of the above-mentioned machine tool prediction method are implemented.

[0043] In addition, to achieve the above-mentioned purpose, the present application also provides a computer-readable storage medium, on which a machine tool prediction program is stored. When the machine tool prediction program is executed by a processor, the steps of the above-mentioned machine tool prediction method are implemented.

[0044] First, the present application can obtain the operating data of the CNC machine tool in real time based on the sampling frequency of the CNC machine tool, wherein the operating data at least includes the temperature data, pressure data and vibration data of the CNC machine tool at the current moment, and the sampling frequency is used to determine the time interval between the current moment and the next moment; that is, the present application collects operating data based on the sampling frequency, while ensuring the quality of the operating data, it also effectively saves the cost and resources of data collection. Then determine the student model specified by the CNC machine tool. The student model refers to a model that is trained and learned from the already trained teacher model based on a preset knowledge distillation algorithm, and both the student model and the teacher model are state detection models of the CNC machine tool. The student model of this application is a relatively simple algorithm model with high computational efficiency and is suitable for real-time monitoring of the operating status of the CNC machine tool. The operating data is then input into the student model and the teacher model respectively, and the actual predicted state output by the student model is constrained according to the given real state output by the teacher model. The real operating state of the CNC machine tool can be accurately determined, thereby effectively avoiding the distortion of the predicted state of the machine tool due to the constraints of predetermined thresholds or rules, thereby improving the monitoring efficiency of the CNC machine tool while also achieving the purpose of improving the accuracy of predicting the operating state of the machine tool. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] FIG1 is a flow chart of a first embodiment of a machine tool prediction method of the present application;

[0046] FIG2 is a flow chart of a second embodiment of the machine tool prediction method of the present application;

[0047] FIG3 is a schematic diagram of the structure of a machine tool prediction device according to an embodiment of the present application;

[0048] FIG4 is a schematic diagram of the structure of a terminal device involved in an embodiment of the present application;

[0049] FIG5 is a schematic diagram of the structure of a computer-readable storage medium involved in an embodiment of the present application.

[0050] The realization of the objectives, functional features and advantages of this application will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0051] An embodiment of the present application provides a machine tool prediction method, as shown in FIG1 , which is a flow chart of a first embodiment of the machine tool prediction method of the present application.

[0052] Exemplary embodiments are described in detail herein, with examples illustrated in the accompanying drawings. When the following description refers to the drawings, identical numerals in different drawings represent identical or similar elements unless otherwise indicated. The embodiments described in the following exemplary embodiments are not intended to represent all embodiments consistent with this application.

[0053] In this embodiment, the machine tool prediction method of the present application is executed by a terminal device that predicts the machine tool operating status, specifically, the control center in the terminal device executes the following machine tool prediction method. In the following embodiments, the execution subject of the present application will not be described in detail.

[0054] The machine tool prediction method of this application includes:

[0055] Step S10: Acquiring operating data of the CNC machine tool based on a sampling frequency of the CNC machine tool, wherein the operating data includes at least temperature data, pressure data, and vibration data of the CNC machine tool at a current moment, and the sampling frequency is used to determine a time interval between the current moment and the next moment;

[0056] In this embodiment, the sampling frequency for data collection of the CNC machine tool is first determined, and then the operation data of the CNC machine tool sent by the machine tool sensor is received according to the sampling frequency, and data preprocessing is performed based on the received operation data to obtain preprocessed operation data (i.e., operation data).

[0057] It should be noted that machine tool sensors include but are not limited to temperature sensors, pressure sensors and vibration sensors; data preprocessing includes but is not limited to cleaning outliers, filling missing values ​​and standardizing values; operating data includes at least the temperature data, pressure data and vibration data of the CNC machine tool at the current moment; the sampling frequency is used to determine the time interval between the current moment and the next moment.

[0058] In one embodiment, after the control center receives the operating data sent by the machine tool sensor, it searches for abnormal data and missing data from the operating data, then deletes these abnormal data from the operating data, and determines the data category of the operating data. Then, the missing data in the operating data is replaced by the average data, median data, or most frequent data (mode) in each data category. Specifically, when the missing data includes at least temperature data, where there are multiple temperature data, the missing temperature value is replaced with the average data, median data, or most frequent data (mode) in the temperature data; then, after completing the cleaning of abnormal values ​​and filling of missing values ​​in the operating data, the operating data is standardized, and the standardized operating data is used as the operating data of the CNC machine tool, thereby effectively improving the training efficiency of data instructions and models.

[0059] Step S20: Determine a student model specified by the CNC machine tool, where the student model is a model that is trained and learned from a pre-trained teacher model according to a preset knowledge distillation algorithm, and both the student model and the teacher model are state detection models of the CNC machine tool;

[0060] In this embodiment, a student model specified by the CNC machine tool is determined, and the student model is trained and learned from a well-trained teacher model through a preset knowledge distillation algorithm, wherein both the student model and the teacher model are state detection models of the CNC machine tool.

[0061] It should be noted that the student model is a relatively simple model, that is, the student model refers to a model that is trained and learned from the already trained teacher model based on the preset knowledge distillation algorithm. The student model has high computational efficiency and is more suitable for real-time monitoring.

[0062] In addition, it should be noted that the teacher model is a deep neural network that has been pre-trained based on the operating data of the CNC machine tool under any working conditions and usage duration, and can capture complex patterns and relationships in the data.

[0063] Step S30: inputting the operating data into the student model and the teacher model respectively, obtaining the actual predicted state of the CNC machine tool output by the student model and the given real state of the CNC machine tool output by the teacher model;

[0064] In this embodiment, the operating data is transmitted to the teacher model to obtain a given real state of the CNC machine tool output by the teacher model, wherein the given real state includes at least a given temperature state, a given pressure state, and a given motion state of the CNC machine tool. While transmitting the operating data to the teacher model, the present application also transmits the operating data to the student model, that is, based on the training and learning of the operating data by the student model, the actual predicted state of the CNC machine tool output by the student model can be quickly obtained, thereby effectively improving the monitoring efficiency of the CNC machine tool, wherein the actual predicted state includes at least a temperature predicted state, a pressure predicted state, and a vibration predicted state of the CNC machine tool.

[0065] Step S40: determining the actual operating state of the CNC machine tool according to the actual predicted state and the given real state.

[0066] In this embodiment, when the actual predicted state is the same as the given real state, the actual predicted state is used as the real operating state of the CNC machine tool; when the actual predicted state is different from the given real state, the prediction error data between the actual predicted state and the given real state is obtained, and the matching result between the actual predicted state and the given real state is determined based on the prediction error data, and the real operating state of the CNC machine tool is determined based on the matching result and the actual predicted state.

[0067] In one embodiment, the present application uses the output value of the teacher model (i.e., the given real state) as the true value of the operating state of the CNC machine tool to constrain the process of inputting the operating data into the student model for training and learning, thereby avoiding the distortion of the data output by the student model, that is, constraining the actual predicted state based on the given real state. Specifically, by comparing the actual predicted state with the given real state, the true operating state of the CNC machine tool can be accurately determined, thereby effectively improving the accuracy of predicting the operating state of the machine tool.

[0068] In summary, the present application can first obtain the operating data of the CNC machine tool in real time based on the sampling frequency of the CNC machine tool, wherein the operating data at least includes the temperature data, pressure data and vibration data of the CNC machine tool at the current moment, and the sampling frequency is used to determine the time interval between the current moment and the next moment; that is, the present application collects operating data based on the sampling frequency, while ensuring the quality of the operating data, it also effectively saves the cost and resources of data collection. Then determine the student model specified by the CNC machine tool. The student model refers to a model that is trained and learned from the already trained teacher model based on a preset knowledge distillation algorithm, and both the student model and the teacher model are state detection models of the CNC machine tool. The student model of this application is a relatively simple algorithm model with high computational efficiency and is suitable for real-time monitoring of the operating status of the CNC machine tool. The operating data is then input into the student model and the teacher model respectively, and the actual predicted state output by the student model is constrained according to the given real state output by the teacher model. The real operating state of the CNC machine tool can be accurately determined, thereby effectively avoiding the distortion of the predicted state of the machine tool due to the constraints of predetermined thresholds or rules, thereby improving the monitoring efficiency of the CNC machine tool while also achieving the purpose of improving the accuracy of predicting the operating state of the machine tool.

[0069] Based on the first embodiment of the machine tool prediction method of the present application, a second embodiment of the machine tool prediction method of the present application is proposed. Referring to FIG. 2 , FIG. 2 is a flow chart of the second embodiment of the machine tool prediction method of the present application.

[0070] In one embodiment, the twin module includes a learning algorithm unit, and the above step S40: determining the actual operating state of the CNC machine tool based on the actual predicted state and the given real state may include the following implementation steps:

[0071] Step S401: when the actual predicted state is the same as the given real state, taking the actual predicted state as the real operating state of the CNC machine tool;

[0072] In this embodiment, it is determined whether the actual predicted state output by the student module is the same as the given real state output by the teacher module. When the actual predicted state output by the student module is the same as the given real state, the actual predicted state is used as the real operating state of the CNC machine tool. It can be determined that the prediction performance of the student model is in a stable state at this time (that is, the prediction performance of the student model is good and very stable and reliable at this time), so that it can be determined that the actual predicted state output by the student module at this time can accurately reflect the real operating state of the CNC machine tool.

[0073] In one embodiment, when the control center obtains that the prediction performance of the student model is in a stable state, the data acquisition frequency specified by the sampling frequency is determined, and the data acquisition frequency is responded to in accordance with the preset frequency reduction instruction, the adjusted sampling frequency is determined, and then the step of obtaining the operation data of the CNC machine tool based on the sampling frequency of the CNC machine tool is returned to according to the adjusted sampling frequency. Specifically, if the data acquisition frequency specified by the sampling frequency is to collect the operation data of the CNC machine tool once every 1 minute, and the prediction performance of the student model is in a stable state, the data acquisition frequency is replaced from collecting the operation data once every 1 minute to collecting the operation data once every 2 minutes in accordance with the preset frequency reduction instruction. At the same time, the acquisition accuracy of the machine tool sensor can be reduced, thereby effectively ensuring the quality of the data and effectively saving the cost and resources of data acquisition.

[0074] Step S402: When the actual predicted state is different from the given real state, obtain prediction error data between the actual predicted state and the given real state, determine the matching result between the actual predicted state and the given real state based on the prediction error data, and determine the real operating state of the CNC machine tool based on the matching result and the actual predicted state.

[0075] In this embodiment, when the actual predicted state is different from the given real state, the prediction error data between the actual predicted state and the given real state can be obtained in a timely manner, and then the matching result between the actual predicted state and the given real state can be quickly determined based on the prediction error data, and the real operating state of the CNC machine tool can be accurately determined based on the matching result and the actual predicted state, thereby further improving the accuracy of predicting the operating state of the machine tool.

[0076] In one embodiment, the actual predicted state includes at least a predicted temperature state, a predicted pressure state, and a predicted vibration state of the CNC machine tool, and the given real state includes at least a given temperature state, a given pressure state, and a given vibration state of the CNC machine tool. In the above step S402 of obtaining prediction error data between the actual predicted state and the given real state, the following implementation steps may also be included:

[0077] Step S4021: determining a temperature error value between the predicted temperature state and the given temperature state, a pressure error value between the predicted pressure state and the given pressure state, and a vibration error value between the predicted vibration state and the given vibration state;

[0078] In this embodiment, when the actual predicted state is different from the given real state, the state type specified by the actual predicted state is determined, wherein the state type includes at least a temperature type, a pressure type, and a vibration type. Then, according to the state type, the given temperature state pointed to by the temperature type, the given pressure state pointed to by the pressure type, and the given vibration state pointed to by the vibration type can be quickly found from the given real state; finally, the temperature error value between the temperature predicted state and the given temperature state, the pressure error value between the pressure predicted state and the given pressure state, and the vibration error value between the vibration predicted state and the given vibration state can be accurately obtained.

[0079] It should be noted that the actual predicted state includes at least the temperature predicted state, pressure predicted state and vibration predicted state of the CNC machine tool; the given real state includes at least the given temperature state, given pressure state and given vibration state of the CNC machine tool.

[0080] Step S4022: determining the temperature mean square deviation of the CNC machine tool according to the temperature error value, determining the pressure mean square deviation of the CNC machine tool according to the pressure error value, and determining the vibration mean square deviation of the CNC machine tool according to the vibration error value;

[0081] In this embodiment, while the temperature mean square deviation of the CNC machine tool can be accurately determined based on the temperature error value, the pressure mean square deviation of the CNC machine tool can also be accurately determined based on the pressure error value, and the vibration mean square deviation of the CNC machine tool can also be accurately determined based on the vibration error value.

[0082] S4023: Superimpose the temperature mean square error, the pressure mean square error, and the vibration mean square error in sequence to obtain prediction error data.

[0083] In this embodiment, the temperature mean square error, the pressure mean square error, and the vibration mean square error are summed up in sequence to obtain prediction error data between the actual predicted state and the given true state.

[0084] In one embodiment, the prediction error data between the actual predicted state and the given true state can be calculated by the following formula (1). Specifically, formula (1) is a loss function, which is expressed as: L = Σ(T(x) - S(x))^2

[0085] Where T(x) represents the given true state output by the teacher model, S(x) represents the actual predicted state output by the student model, x represents the operating data sent by the machine tool sensor based on the sampling frequency of the CNC machine tool, Σ represents the summation operation, and L represents the prediction error data.

[0086] In one embodiment, in step S402, determining a matching result between the actual predicted state and the given real state based on the prediction error data may include the following implementation steps:

[0087] Step S4024: determining an error threshold range specified for the given real state, and detecting whether the predicted error data exceeds the error threshold range;

[0088] In this embodiment, while determining the error threshold range specified for a given real state, it is immediately detected whether the predicted error data exceeds the error threshold range.

[0089] Step S4025: If the prediction error data does not exceed the error threshold range, determining that the matching result between the actual prediction state and the given real state is that the actual prediction state matches the given real state;

[0090] In this embodiment, if the prediction error data does not exceed the error threshold range, the matching result between the actual predicted state and the given true state is determined to be that the actual predicted state matches the given true state, so that it can be quickly determined that the prediction performance of the student model at this time is general, that is, there is a certain error between the actual predicted state output by the student model at this time and the given true state, but the actual predicted state output by the student model at this time still has a certain reference value.

[0091] Step S4026: If the prediction error data exceeds the error threshold range, determining that the matching result between the actual prediction state and the given real state is that the actual prediction state does not match the given real state.

[0092] In this embodiment, if the prediction error data exceeds the error threshold range, the matching result between the actual predicted state and the given real state is determined to be that the actual predicted state does not match the given real state, so that it can be quickly determined that the prediction performance of the student model at this time is poor, that is, the actual predicted state output by the student model at this time is distorted and has no reference value.

[0093] In one embodiment, the above step S402 of determining the actual operating state of the CNC machine tool based on the matching result and the actual predicted state may further include the following implementation steps:

[0094] Step S4027: When the matching result is that the actual predicted state matches the given true state, the student model is updated according to the loss function specified by the prediction error data to obtain an updated student model, and the step of inputting the operating data into the student model and the teacher model is returned to be executed according to the updated student model, until a target predicted state identical to the given true state is obtained through training, and the target predicted state is used as the true operating state of the CNC machine tool;

[0095] In this embodiment, when the matching result is that the actual predicted state matches the given real state, the time identifier pointed to by the actual predicted state output by the student model at this time is determined. When the time node specified by the time identifier is the same as the update time corresponding to the student model, the loss function specified by the prediction error data is determined, and the student model is updated according to the loss function to obtain an updated student model, thereby achieving the purpose of continuously updating the student model so that the student model can automatically adapt to the changes in the CNC machine tool, thereby effectively improving the generalization ability and stability of the student model. The updated student model is then used as the next student model, and the step of inputting the operating data into the student model and the teacher model is returned to execute, until the target predicted state that is the same as the given real state is obtained after training, and the target predicted state is used as the real operating state of the CNC machine tool, thereby effectively improving the accuracy of predicting the operating state of the machine tool.

[0096] It should be noted that the target prediction state refers to the actual prediction state output by the final updated student model. The change of CNC machine tools can be understood as the wear degree of machine tool components may change with the use of CNC machine tools, and the change of CNC machine tools will directly affect the operating status of the machine tools.

[0097] Step S4028: When the matching result indicates that the actual predicted state does not match the given real state, the sampling frequency is adjusted according to the actual predicted state to obtain an adjusted sampling frequency, and the step of obtaining the operating data of the CNC machine tool based on the sampling frequency of the CNC machine tool is returned to according to the adjusted sampling frequency;

[0098] In this embodiment, when the matching result is that the actual predicted state does not match the given real state, the actual predicted state is determined to be distorted data, that is, the prediction performance of the student model at this time is in a failed state (that is, an unstable state), and the frequency increase instruction of the CNC machine tool is triggered based on the distorted data, and the sampling frequency increment specified by the frequency increase instruction is determined. The sampling frequency of the CNC machine tool is automatically adjusted based on the sampling frequency increment, that is, the sampling frequency increment is superimposed on the sampling frequency to obtain the adjusted sampling frequency, thereby effectively improving the prediction performance of the student model, and then the sampling frequency based on the CNC machine tool is returned to execute based on the adjusted sampling frequency to obtain the operation data of the CNC machine tool.

[0099] In one embodiment, if the sampling frequency specifies that the data collection frequency is to collect the operating data of the CNC machine tool once every 1 minute, and the predictive performance of the student model is in a failed state, the sampling frequency of the CNC machine tool is adaptively adjusted based on the sampling frequency increment to adjust the data collection frequency from collecting operating data once every 1 minute to collecting operating data once every 30 seconds, while also increasing the acquisition accuracy of the machine tool sensor. That is, the present application determines based on the distorted data that the predictive performance of the student model is in a failed state and automatically adjusts the sampling frequency of the CNC machine tool, thereby dynamically adjusting the sampling frequency of the CNC machine tool. This adaptive data sampling strategy effectively saves data acquisition costs and resources while ensuring data quality.

[0100] In one embodiment, the above step S4027: updating the student model according to the loss function specified by the prediction error data may further include the following implementation steps:

[0101] Step S40271: determining the gradient data of the CNC machine tool according to the loss function specified by the prediction error data, and determining the learning frequency specified by the student model;

[0102] In this embodiment, the loss function specified by the prediction error data can quickly determine the gradient data of the CNC machine tool. The gradient data refers to the derivative of the loss function, and then determines the learning frequency specified by the student model. Different prediction error data corresponds to different loss functions.

[0103] Step S40272: Update the student model according to the product data between the gradient data and the learning frequency to obtain an updated student model.

[0104] In this embodiment, the student model is updated according to the product data between the gradient data and the learning frequency to obtain an updated student model.

[0105] In one embodiment, the present application may also update the student model according to a preset incremental learning algorithm to obtain an updated student model, wherein the incremental learning algorithm is shown in the following formula (2):

[0106] Among them, S' is the updated student model, S is the original student model, η is the learning rate, is the gradient data.

[0107] In one embodiment, the machine tool prediction method may further include the following implementation steps:

[0108] Step A10: constructing a digital twin model of the CNC machine tool based on the equipment data of the CNC machine tool, transmitting the operating data to the digital twin model, and outputting the actual operating status of the CNC machine tool, wherein the equipment data includes at least the operating data, machine tool tool wear data, and machine tool workpiece data;

[0109] In this embodiment, a digital twin model of the CNC machine tool is constructed based on the equipment data of the CNC machine tool, and the operation data is transmitted to the digital twin model to output the actual operation status of the CNC machine tool, wherein the equipment data includes at least machine tool status data, machine tool tool wear data and machine tool workpiece data.

[0110] It should be noted that the machine tool status data at least includes the historical operating status; the machine tool tool wear data at least includes the tool cutting parameters, tool life parameters and tool pressure parameters; the machine tool workpiece data at least includes the geometric structure parameters and processing quality parameters of the processed workpiece; the digital twin model is used to reflect the actual operating status of the CNC machine tool, that is, the digital twin model is a virtual entity of the physical entity of the CNC machine tool.

[0111] Step A20: After the digital twin model obtains the real operating state, it is determined whether the actual operating state is consistent with the real operating state;

[0112] In this embodiment, after the digital twin model obtains the actual operating status, the digital twin model is used to determine whether the actual operating status is consistent with the actual operating status, that is, the digital twin model is used to determine whether the actual operating status is the same as the actual operating status, or / and the digital twin model is used to determine whether the actual operating status is within the status threshold range corresponding to the actual operating status.

[0113] Step A30: If the actual operating state is consistent with the true operating state, determining that the CNC machine tool is in a normal operating state;

[0114] In this embodiment, if the actual operating state is consistent with the real operating state, that is, the actual operating state is the same as the real operating state, or / and the actual operating state is within the state threshold range corresponding to the real operating state, then it can be directly determined that the CNC machine tool is in a normal operating state, and the normal operating state can be transmitted to the human-computer interaction interface of the CNC machine tool through the digital twin model for display.

[0115] Step A40: If the actual operating state is inconsistent with the true operating state, it is determined that the CNC machine tool is in an abnormal operating state, and a fault prompt is given according to the abnormal operating state.

[0116] In this embodiment, if the actual operating state is inconsistent with the real operating state, that is, the actual operating state is different from the real operating state, or / and the actual operating state is not within the state threshold range corresponding to the real operating state, then it is determined that the CNC machine tool is in an abnormal operating state, and a fault prompt is transmitted to the human-computer interaction interface of the CNC machine tool through the digital twin model based on the abnormal operating state, and the real operating state is adjusted to be the same as the actual operating state. That is, this application not only achieves the purpose of real-time monitoring of the machine tool operating state, but also effectively improves the early warning and prevention capabilities of machine tool failures.

[0117] In summary, the student model of the present application refers to a model that is trained and learned from a trained teacher model based on a preset knowledge distillation algorithm, thereby realizing real-time, accurate and efficient prediction of the machine tool operating status. When the actual predicted state matches the given true state, the student model is updated according to the loss function specified by the prediction error data, thereby achieving the purpose of continuously updating the student model, so that the student model can automatically adapt to the changes in the CNC machine tool, thereby effectively improving the generalization ability and stability of the student model; when the actual predicted state does not match the given true state or / and when the actual predicted state is the same as the given true state, the sampling frequency of the CNC machine tool is adaptively adjusted according to the prediction performance of the student model, while ensuring data quality, achieving the purpose of saving data acquisition costs and resources; then the control center sends the actual predicted state output by the final updated student model to the digital twin model, and compares it with the actual operating state output by the digital twin model, so that the control center can adjust the machine tool operating state in real time, thereby effectively improving the early warning and prevention capabilities of machine tool failures.

[0118] In addition, the present application also provides a machine tool prediction device. Please refer to Figure 3, which is a structural diagram of the machine tool prediction device involved in the embodiment of the present application.

[0119] The machine tool prediction device of this application includes:

[0120] an acquisition module H01 for acquiring operating data of a CNC machine tool based on a sampling frequency of the CNC machine tool, wherein the operating data includes at least temperature data, pressure data, and vibration data of the CNC machine tool at a current moment, and the sampling frequency is used to determine a time interval between the current moment and the next moment;

[0121] A designation module H02 is configured to determine a student model designated by the CNC machine tool, wherein the student model is a model that is trained to learn from a pre-trained teacher model according to a preset knowledge distillation algorithm, and both the student model and the teacher model are state detection models of the CNC machine tool;

[0122] A model output module H03 inputs the operating data into the student model and the teacher model respectively, and obtains the actual predicted state of the CNC machine tool output by the student model and the given real state of the CNC machine tool output by the teacher model;

[0123] The prediction module H04 determines the actual operating state of the CNC machine tool according to the actual predicted state and the given real state.

[0124] In one embodiment, the prediction module H04 further includes:

[0125] a first comparing unit, configured to use the actual predicted state as the actual operating state of the CNC machine tool when the actual predicted state is the same as the given actual state;

[0126] A first matching unit is used to obtain prediction error data between the actual predicted state and the given real state when the actual predicted state is different from the given real state, determine a matching result between the actual predicted state and the given real state based on the prediction error data, and determine the real operating state of the CNC machine tool based on the matching result and the actual predicted state.

[0127] In one embodiment, the prediction module H04 further includes:

[0128] an error determination unit, configured to determine a temperature error value between the predicted temperature state and the given temperature state, a pressure error value between the predicted pressure state and the given pressure state, and a vibration error value between the predicted vibration state and the given vibration state;

[0129] a mean square deviation determining unit, configured to determine a temperature mean square deviation of the CNC machine tool based on the temperature error value, determine a pressure mean square deviation of the CNC machine tool based on the pressure error value, and determine a vibration mean square deviation of the CNC machine tool based on the vibration error value;

[0130] The superposition unit is used to superimpose the temperature mean square error, the pressure mean square error and the vibration mean square error in sequence to obtain prediction error data.

[0131] In one embodiment, the prediction module H04 further includes:

[0132] a detection unit, configured to determine an error threshold range specified for the given real state, and detect whether the predicted error data exceeds the error threshold range;

[0133] a second matching unit, configured to determine, if the prediction error data does not exceed the error threshold range, that a matching result between the actual predicted state and the given real state is that the actual predicted state matches the given real state;

[0134] The third matching unit is configured to determine, if the prediction error data exceeds the error threshold range, that a matching result between the actual predicted state and the given real state is that the actual predicted state does not match the given real state.

[0135] In one embodiment, the prediction module H04 further includes:

[0136] a training unit, configured to, when the matching result is that the actual predicted state matches the given true state, update the student model according to a loss function specified by the prediction error data to obtain an updated student model, and return to executing the step of inputting the operating data into the student model and the teacher model respectively according to the updated student model, until a target predicted state identical to the given true state is obtained through training, and use the target predicted state as the true operating state of the CNC machine tool;

[0137] An adjustment unit is used to adjust the sampling frequency according to the actual predicted state when the matching result is that the actual predicted state does not match the given real state, to obtain an adjusted sampling frequency, and return to execute the step of obtaining the operating data of the CNC machine tool based on the sampling frequency of the CNC machine tool according to the adjusted sampling frequency.

[0138] In one embodiment, the prediction module H04 further includes:

[0139] a specifying unit, configured to determine the gradient data of the CNC machine tool according to the loss function specified by the prediction error data, and determine the learning frequency specified by the student model;

[0140] An updating unit is used to update the student model according to the product data between the gradient data and the learning frequency to obtain an updated student model.

[0141] In one embodiment, the prediction module H04 further includes:

[0142] a transmission unit, configured to construct a digital twin model of the CNC machine tool based on the equipment data of the CNC machine tool, transmit the operation data to the digital twin model, and output the actual operation status of the CNC machine tool, wherein the equipment data includes at least machine status data, machine tool tool wear data, and machine workpiece data;

[0143] a judgment unit, configured to judge whether the actual operating state is consistent with the real operating state after the digital twin model obtains the real operating state;

[0144] a second comparing unit, configured to determine that the CNC machine tool is in a normal operating state if the actual operating state is consistent with the true operating state;

[0145] A fault prompt unit is used to determine that the CNC machine tool is in an abnormal operating state if the actual operating state is inconsistent with the real operating state, and to provide a fault prompt based on the abnormal operating state.

[0146] Each functional module of the machine tool prediction device of the present application implements the steps of the machine tool prediction method of the present application as described above during operation.

[0147] In addition, the present application also provides a terminal device. Please refer to Figure 4, which is a schematic diagram of the structure of the terminal device involved in the embodiment of the present application. The terminal device in the embodiment of the present application can specifically be a device for predicting the operation of a local machine tool.

[0148] As shown in Figure 4, the terminal device of the embodiment of the present application may include: a processor 1001, such as a CPU, a communication bus 1002, a user interface 1003, a network interface 1004, and a memory 1005. Among them, the communication bus 1002 is used to realize the connection and communication between these components. The user interface 1003 may include a display screen (Display), an input unit such as a keyboard (Keyboard), and the user interface 1003 may also include a standard wired interface and a wireless interface. The network interface 1004 may include a standard wired interface and a wireless interface (such as a Wi-Fi interface).

[0149] Memory 1005 is provided on the terminal device body. Memory 1005 stores programs that, when executed by processor 1001, implement corresponding operations. Memory 1005 is also used to store parameters used by the terminal device. Memory 1005 can be high-speed RAM memory or non-volatile memory, such as disk storage. Memory 1005 can also be a storage device independent of processor 1001.

[0150] Those skilled in the art will understand that the terminal device structure shown in FIG4 does not constitute a limitation on the terminal device, and may include more or fewer components than shown in the figure, or combine certain components, or arrange the components differently.

[0151] As shown in FIG. 4 , the memory 1005 as a storage medium may include an operating system, a network communication module, a user interface module, and a machine tool prediction program of a terminal device.

[0152] In the terminal device shown in FIG4 , the processor 1001 may be configured to call a machine tool prediction program of the terminal device stored in the memory 1005 and execute the steps of the above-mentioned machine tool prediction method.

[0153] In addition, please refer to Figure 5, which is a schematic diagram of the structure of a computer-readable storage medium involved in an embodiment of the present application. This application provides a computer-readable storage medium. The computer-readable storage medium stores a machine tool prediction program, which, when executed by a processor, implements the steps of the above-described machine tool prediction method.

[0154] It should be noted that, in this document, the terms "comprises," "includes," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, method, article, or system comprising a series of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or system. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of other identical elements in the process, method, article, or system comprising the element.

[0155] The serial numbers of the above embodiments of the present application are for description only and do not represent the advantages or disadvantages of the embodiments.

[0156] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus the necessary general hardware platform, and of course can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art can be embodied in the form of a software product, which is stored in a computer-readable storage medium (such as ROM / RAM, magnetic disk, optical disk) as mentioned above, and includes a number of instructions for enabling a terminal device (which can be a mobile phone, computer, server, or network device, etc.) to execute the methods described in each embodiment of the present application.

[0157] The above are merely embodiments of the present application and are not intended to limit the patent scope of the present application. Any equivalent structure or equivalent process transformation made using the contents of the present application specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present application.

Claims

1. A machine tool prediction method, wherein: The machine tool prediction method comprises: Acquire operation data of the CNC machine tool based on a sampling frequency of the CNC machine tool, wherein the operation data at least includes temperature data, pressure data and vibration data of the CNC machine tool at a current moment, and the sampling frequency is used to determine a time interval between the current moment and a next moment; Determine a student model specified by the CNC machine tool, wherein the student model refers to a model that is trained and learned from a trained teacher model according to a preset knowledge distillation algorithm, and both the student model and the teacher model are state detection models of the CNC machine tool; Inputting the operating data into the student model and the teacher model respectively, obtaining the actual predicted state of the CNC machine tool output by the student model and the given real state of the CNC machine tool output by the teacher model; The actual operating state of the numerically controlled machine tool is determined according to the actual predicted state and the given real state.

2. The machine tool prediction method according to claim 1, wherein: The step of determining the actual operating state of the CNC machine tool according to the actual predicted state and the given real state comprises: When the actual predicted state is the same as the given real state, taking the actual predicted state as the real operating state of the CNC machine tool; When the actual predicted state is different from the given real state, prediction error data between the actual predicted state and the given real state is obtained, a matching result between the actual predicted state and the given real state is determined based on the prediction error data, and the real operating state of the CNC machine tool is determined based on the matching result and the actual predicted state.

3. The machine tool prediction method according to claim 2, wherein: The actual predicted state includes at least the temperature predicted state, pressure predicted state and vibration predicted state of the CNC machine tool, and the given real state includes at least the given temperature state, given pressure state and given vibration state of the CNC machine tool; The step of obtaining prediction error data between the actual predicted state and the given true state comprises: Determining a temperature error value between the temperature prediction state and the given temperature state, a pressure error value between the pressure prediction state and the given pressure state, and a vibration error value between the vibration prediction state and the given vibration state; Determine the temperature mean square deviation of the CNC machine tool according to the temperature error value, determine the pressure mean square deviation of the CNC machine tool according to the pressure error value, and determine the vibration mean square deviation of the CNC machine tool according to the vibration error value; The temperature mean square error, the pressure mean square error and the vibration mean square error are superimposed in sequence to obtain prediction error data.

4. The machine tool prediction method according to claim 3, wherein: The step of determining the matching result between the actual predicted state and the given real state according to the prediction error data comprises: Determining an error threshold range specified by the given real state, and detecting whether the predicted error data exceeds the error threshold range; If the prediction error data does not exceed the error threshold range, determining that the matching result between the actual prediction state and the given real state is that the actual prediction state matches the given real state; If the prediction error data exceeds the error threshold range, it is determined that the matching result between the actual prediction state and the given real state is that the actual prediction state does not match the given real state.

5. The machine tool prediction method according to claim 4, wherein: The step of determining the actual operating state of the CNC machine tool according to the matching result and the actual predicted state comprises: When the matching result is that the actual predicted state matches the given real state, the student model is updated according to the loss function specified by the prediction error data to obtain an updated student model, and the step of inputting the operating data into the student model and the teacher model is returned to execute according to the updated student model, until a target predicted state identical to the given real state is obtained through training, and the target predicted state is used as the real operating state of the CNC machine tool; When the matching result is that the actual predicted state does not match the given real state, the sampling frequency is adjusted according to the actual predicted state to obtain the adjusted sampling frequency, and the step of obtaining the operating data of the CNC machine tool based on the sampling frequency of the CNC machine tool is returned according to the adjusted sampling frequency.

6. The machine tool prediction method according to claim 5, wherein: The step of updating the student model according to the loss function specified by the prediction error data comprises: Determining the gradient data of the numerically controlled machine tool according to the loss function specified by the prediction error data, and determining the learning frequency specified by the student model; The student model is updated according to the product data between the gradient data and the learning frequency to obtain an updated student model.

7. The machine tool prediction method according to any one of claims 1 to 6, wherein: The machine tool prediction method further comprises: Building a digital twin model of the CNC machine tool based on the equipment data of the CNC machine tool, transmitting the operation data to the digital twin model, and outputting the actual operation status of the CNC machine tool, wherein the equipment data at least includes machine status data, machine tool tool wear data, and machine workpiece data; After the digital twin model obtains the real operating state, determining whether the actual operating state is consistent with the real operating state; If the actual operating state is consistent with the real operating state, it is determined that the CNC machine tool is in a normal operating state; If the actual operating state is inconsistent with the true operating state, it is determined that the CNC machine tool is in an abnormal operating state, and a fault prompt is given according to the abnormal operating state.

8. A machine tool prediction device, wherein: The machine tool prediction device comprises: an acquisition module, configured to acquire operation data of the CNC machine tool based on a sampling frequency of the CNC machine tool, wherein the operation data at least includes temperature data, pressure data and vibration data of the CNC machine tool at a current moment, and the sampling frequency is used to determine a time interval between the current moment and a next moment; A designation module, used to determine a student model designated by the CNC machine tool, wherein the student model refers to a model that is trained and learned from a trained teacher model according to a preset knowledge distillation algorithm, and both the student model and the teacher model are state detection models of the CNC machine tool; A model output module, inputting the operating data into the student model and the teacher model respectively, to obtain the actual predicted state of the CNC machine tool output by the student model and the given real state of the CNC machine tool output by the teacher model; The prediction module determines the actual operating state of the CNC machine tool according to the actual predicted state and the given real state.

9. A terminal device, wherein: The terminal device includes a memory, a processor, and a machine tool prediction program stored in the memory and executable on the processor. When the processor executes the machine tool prediction program, the steps of the machine tool prediction method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium, wherein: The computer-readable storage medium stores a machine tool prediction program, which, when executed by a processor, implements the steps of the machine tool prediction method according to any one of claims 1 to 7.

Images