TWI900033B - Thermal displacement adaptive computer numerical control compensation system and method based on meta-learning - Google Patents

Abstract

The present invention relates to a thermal displacement adaptive computer numerical control compensation system and method based on meta-learning. The system involves an equipment communication module, configured to receive an equipment data and write a compensation value to an equipment; a temperature communication module, configured to receive a temperature data; a displacement communication module, configured to receive a displacement data; a signal processing unit, connected to the above-mentioned modules to process and store the above-mentioned data; a diagnosis module, connected to the signal processing unit, performing evaluation and diagnosis based on the above-mentioned data to build a thermal displacement compensation model; a deployment module, connected to the signal processing unit and using a deep learning model to output a prediction value; and a host system communication module, connected to the signal processing unit, and supporting many-to-one communication connections to connect to a host system that performs a meta learning.

Description

Thermal displacement adaptive computer numerical control compensation system and method based on meta-learning

The present invention relates to a computer numerical control compensation system and method, and more particularly to a meta-learning-based thermal displacement adaptive computer numerical control compensation system and method.

In early thermal compensation methods, users typically relied on their processing experience to determine downtime, meaning they would wait until the machine temperature returned to the ideal range before continuing processing to ensure that the product met quality standards.

The current strategy is to install temperature sensors on key components of the equipment. This involves long periods of downtime at different seasonal temperatures to collect data. Data analytics experts then conduct feature analysis, develop thermal compensation models, and ultimately implement compensation.

However, these conventional methods not only have limited effectiveness in improving machining accuracy but also involve additional time and cost. Consequently, these methods fail to effectively address the aforementioned accuracy and cost issues, and therefore have limited market acceptance.

The purpose of this invention is to provide a meta-learning-based adaptive computer numerical control compensation system for thermal displacement that can reduce downtime, is not restricted by ambient temperature, and reduces modeling costs. The system can automatically collect temperature and displacement data while also generating heat energy during equipment operation or processing.

Another object of the present invention is to provide a meta-learning-based adaptive computer numerical control compensation system for thermal displacement. By using meta-learning technology, a small amount of data and data splicing can be used to overcome the problem of insufficient data and quickly establish a thermal compensation model.

Another object of the present invention is to provide a meta-learning-based adaptive thermal displacement computer numerical control compensation system that stores temperature data, displacement data, equipment data, and compensation values in a host system such as a cloud server (e.g., the host system's information database) for user query and analysis of historical data. The system also supports remote updating of its thermal compensation model, improving compensation accuracy through model learning.

In accordance with the above-mentioned objectives, the present invention provides a meta-learning-based thermal displacement adaptive computer numerical control compensation system, comprising: a device communication module, configured to be electrically connected to a controller of a corresponding device to receive device data of the device and write a compensation value to the device for compensation; a temperature communication module, configured to be electrically connected to at least one temperature sensor located at a predetermined position of the device to receive temperature data read and collected by the at least one temperature sensor; a displacement communication module, configured to be electrically connected to a displacement sensor located on the device to receive displacement data read and collected by the displacement sensor; and a signal processing unit. , configured to be electrically connected to the device communication module, the temperature communication module, and the displacement communication module to process and store the device data, the temperature data, and the displacement data; a diagnosis module, configured to be electrically connected to the signal processing unit, to perform evaluation and diagnosis based on the device data, the temperature data, and the displacement data in a predetermined manner to establish a thermal displacement compensation model; a deployment module, configured to be electrically connected to the signal processing unit, to use a deep learning model to output a prediction value as a compensation benchmark; and a host system communication module, configured to be electrically connected to the signal processing unit, to support many-to-one communication connections and to execute a host system of unary learning.

In some embodiments, the meta-learning-based thermal displacement adaptive computer numerical control compensation system further includes an interface display that displays the temperature data, the compensation value, and the displacement data in a corresponding graph through internal software, and is configured to set a compensation mechanism according to user requirements and upload the temperature data, the compensation value, and the displacement data to the host system.

In some embodiments, the predetermined method includes at least residual analysis and overfitting.

In some embodiments, the equipment data includes at least the operating status and processing time of the equipment itself.

In some embodiments, the predetermined location is inside the device, outside the device, or the location of a key component of the device.

In some embodiments, the signal processing unit is a microprocessor having a physical transmission interface of one or more of electrical potential transmission function, Ethernet or Wi-Fi, and input/output (IO) communication.

In some embodiments, the host system is a cloud server.

In some embodiments, the host system includes an original model module that performs feature engineering on the temperature data and then establishes a basic compensation model.

In some embodiments, the upper system further includes an update module that provides the base compensation model with meta-learning and data splicing for updating, and then returns the updated base compensation model to the update module via the deployment module, providing a data analysis and model training function that is relatively unfamiliar to the client.

In some embodiments, the upper system further includes a meta-learning module, and the update module updates the basic compensation model by performing the meta-learning and data splicing through the meta-learning module. The meta-learning module uses a few-shot learning method to learn new tasks using a small amount of data.

The present invention provides another meta-learning-based thermal displacement adaptive computer numerical control compensation method, comprising: collecting temperature data and displacement data; storing the temperature data and the displacement data and uploading them to a cloud server; performing feature analysis on the temperature data and the displacement data; establishing a basic compensation model; and performing thermal error compensation on a device and returning to the step of collecting the temperature data and the displacement data. The step of establishing the basic compensation model simultaneously includes remotely updating the basic compensation model and performing meta-learning and data splicing to update it.

In some embodiments, before collecting the temperature data and the displacement data, the method further includes installing a temperature sensor at a predetermined location of the device and confirming communication of the temperature sensor. The predetermined location is inside the device, outside the device, or at a key component of the device.

In some embodiments, before the step of collecting the temperature data and the displacement data and after installing the temperature sensor at the predetermined location of the device and confirming the communication of the temperature sensor, it also includes installing a displacement sensor on the device and confirming the communication of the displacement sensor.

In some embodiments, before collecting the temperature data and the displacement data and after installing the displacement sensor on the device and confirming the communication of the displacement sensor, the method further includes performing measurement using a standard rod.

In order to make the above-mentioned objects, features and advantages of the present invention more clearly understood, the following describes in detail the specific embodiments listed in the accompanying drawings.

The advantages, features, and technical methods achieved by the present invention will be described in more detail with reference to exemplary embodiments and the accompanying drawings for easier understanding. The present invention can be implemented in various forms, and therefore should not be construed as being limited to the embodiments described herein. On the contrary, for those skilled in the art, the provided embodiments will enable this disclosure to more thoroughly and comprehensively convey the scope of the present invention, and the present invention will be defined solely by the scope of the appended patent applications.

In addition, the terms "include" and/or "comprising" refer to the existence of the stated features, regions, wholes, steps, operations, elements and/or parts, but do not exclude the existence or addition of one or more other features, regions, wholes, steps, operations, elements, parts and/or combinations thereof.

To help you better understand the content of this invention and the effects it can achieve, the following specific embodiments are described in detail with reference to the accompanying drawings.

Figure 1 is a block diagram of the thermal displacement adaptive computer numerical control compensation system based on meta-learning of the present invention.

1 , the meta-learning-based thermal displacement adaptive computer numerical control compensation system 100 of the present invention includes a device communication module 10, a temperature communication module 20, a displacement communication module 30, a signal processing unit 40, a diagnosis module 50, a deployment module 60, a host system communication module 70, and an interface display 80.

The device communication module 10 is configured to electrically connect to a device controller 210 of a corresponding device 200 to receive device data of the device 200 and write a compensation value to the device 200 for compensation. In some embodiments, the device data includes at least the operating status and processing time of the device 200 itself, but is not limited thereto.

In some embodiments, device 200 is equipped with at least one temperature sensor 220 and one displacement sensor 230. Temperature sensor 220 may be located at a predetermined location on device 200. In some embodiments, the predetermined location may be inside device 200, outside device 200, or at a key component of device 200, but is not limited thereto. Displacement sensor 230 may be located on device 200.

The temperature communication module 20 is configured to be electrically connected to at least one temperature sensor 220 located at the predetermined position of the device 200 to receive temperature data read and collected by the at least one temperature sensor 220 .

The displacement communication module 30 is configured to be electrically connected to the displacement sensor 230 located on the device 200 to receive displacement data read and collected by the displacement sensor 230 .

The signal processing unit 40 is configured to electrically connect to the device communication module 10, the temperature communication module 20, and the displacement communication module 30 to process and store the device data, the temperature data, and the displacement data. In some embodiments, the signal processing unit 40 may be a microprocessor having a physical transmission interface including, but not limited to, voltage transmission, Ethernet or Wi-Fi, or input/output (IO) communication.

The diagnostic module 50 is electrically connected to the signal processing unit 40 and performs evaluation and diagnosis based on the device data, the temperature data, and the displacement data using a predetermined method to establish a thermal displacement compensation model. In some embodiments, the predetermined method includes at least residual analysis and overfitting, but is not limited thereto.

The deployment module 60 is configured to be electrically connected to the signal processing unit 40 and uses a deep learning model to output a prediction value as a compensation benchmark.

FIG2 is a block diagram showing the structure of the host system for multiple thermal displacement adaptive computer numerical control compensation systems based on meta-learning according to the present invention.

1 and 2 , the host system communication module 70 is configured to be electrically connected to the signal processing unit 40 to support many-to-one (as shown in FIG. 2 ) communication connections and to execute a host system 300 for unary learning.

In some embodiments, the meta-learning-based adaptive thermal displacement computer numerical control compensation system 100 of the present invention further includes an interface display 80. The interface display 80 displays the temperature data, the compensation value, and the displacement data in corresponding graphs via internal software (e.g., an application program). The interface display 80 is configured to set a compensation mechanism based on user requirements and upload the temperature data, the compensation value, the displacement data, and other related data to the host system 300.

Referring back to FIG. 1 , in some embodiments, the host system 300 may be a cloud server. In some embodiments, the host system 300 may include an original model module 310 . The original model module 310 may perform feature engineering on the temperature data and then establish a basic compensation model.

In some embodiments, the host system 300 may further include an update module 320. The update module 320 may perform meta-learning and data splicing on the original base compensation model to update it, and then return the updated base compensation model to the update module 320 via the deployment module 60, providing a relatively unfamiliar data analysis and model training function for the client.

In some embodiments, the host system 300 may further include a meta-learning module 330. The update module 320 uses the meta-learning module 330 to update the base compensation model by performing meta-learning and data splicing. The meta-learning module 330 may employ a few-shot learning approach to learn new tasks using a small amount of data. Thus, meta-learning techniques can overcome data shortages by using a small amount of data and data splicing to quickly establish a hot compensation model (e.g., the base compensation model).

Therefore, the meta-learning-based adaptive thermal displacement computer numerical control compensation system 100 can reduce downtime, be unrestricted by ambient temperature, and lower modeling costs. The system automatically collects temperature and displacement data, while also automatically collecting heat energy generated by equipment operation or processing. Furthermore, the system stores temperature data, displacement data, equipment data, and compensation values in a host system 300, such as a cloud server (e.g., a database of the host system), allowing users to query and analyze historical data. The system also supports remote updating of its thermal compensation model (e.g., the basic compensation model), improving compensation accuracy through model learning.

FIG3 is a flow chart of the meta-learning-based thermal displacement adaptive computer numerical control compensation method of the present invention.

Referring to FIG. 1 and FIG. 3 , the meta-learning-based thermal displacement adaptive computer numerical control compensation method S100 of the present invention includes: collecting temperature data and displacement data (step S14); storing the temperature data and the displacement data and uploading them to a cloud server (step S15); performing feature analysis on the temperature data and the displacement data (step S16); establishing a basic compensation model (step S17); and performing thermal error compensation on a device and returning to the step of collecting the temperature data and the displacement data (step S18). Furthermore, while executing the step of building the basic compensation model (step S17), the basic compensation model is synchronously updated remotely (step S191) and a univariate learning and data splicing are performed to update the model (step S192).

In some embodiments, step S14 may collect the temperature data by the temperature sensor 220 and the displacement data by the displacement sensor 230 as shown in FIG. 1 . The detailed description of the temperature sensor 220 and the displacement sensor 230 is the same as that of FIG. 1 , and thus will not be repeated here.

In some embodiments, step S15 can be implemented by the device communication module 10, the temperature communication module 20, the displacement communication module 30, the signal processing unit 40, the diagnosis module 50, the deployment module 60, and the upper system communication module 70 as shown in FIG1 . The detailed description is the same as that of FIG1 above, so it will not be repeated here.

In some embodiments, step S16 may be implemented by the original model module 310 in the host system 300 as shown in FIG1 . The detailed description is the same as that of FIG1 , and thus will not be repeated here.

In some embodiments, step S191 may be implemented by the update module 320 in the host system 300 as shown in FIG1 . The detailed description is the same as that of FIG1 , and thus will not be repeated here.

In some embodiments, step S192 may be implemented by the meta-learning module 330 in the host system 300 as shown in FIG1 . The detailed description is the same as that of FIG1 , and thus will not be repeated here.

Referring again to FIG. 3 , in some embodiments, before collecting the temperature data and the displacement data (i.e., step S14 ), the process further includes installing a temperature sensor at a predetermined location on the device and confirming communication with the temperature sensor (step S11 ). The predetermined location may be inside the device, outside the device, or at a key component of the device.

In some embodiments, before the step of collecting the temperature data and the displacement data (i.e., step S14) and after installing the temperature sensor at the predetermined location of the device and confirming the communication of the temperature sensor (i.e., step S11), it also includes installing a displacement sensor on the device and confirming the communication of the displacement sensor (step S12).

In some embodiments, before collecting the temperature data and the displacement data (i.e., step S14) and after installing the displacement sensor on the device and confirming communication with the displacement sensor (i.e., step S12), the method further includes performing measurements using a calibration rod (step S13). In some embodiments, the calibration rod, which can also be referred to as a calibration blade, is configured to measure thermal displacement (i.e., displacement).

In an actual operation example of thermal displacement adaptation, the following are included: (1) communicating with the device 200 by setting the IP address to obtain the current status of the device 200 (e.g., device data); (2) displaying the current temperature value of at least one temperature sensor 220 (e.g., temperature data), storing the data, and uploading it to the cloud (e.g., the upper system 300 of the cloud server); (3) obtaining the thermal displacement trend through a curve chart, and storing the data and uploading it to the cloud (e.g., the upper system 300 of the cloud server).

Figure 4 is a schematic diagram of data splicing for the meta-learning-based thermal displacement adaptive computer numerical control compensation system of the present invention. Figure 5 is a schematic diagram of the thermal compensation model verification for the meta-learning-based thermal displacement adaptive computer numerical control compensation system of the present invention.

Referring to Figure 4, the host system 300 uses data analysis, meta-learning, and data concatenation to build a model. The predicted and actual values are then verified to obtain a suitable thermal compensation model. In Figure 4, for example, Data 1, Data 2, Data 3, and Data 4 are all temperature and displacement data. Furthermore, each data point represents displacement data generated by temperature changes, but this is not the only limitation.

Referencing Figures 4 and 5, the validation results comparing the predicted values (dashed curve in Figure 5) and the actual values (solid curve in Figure 5) show that they become increasingly close after repeated model training and updates. Therefore, training and updating to improve model accuracy not only reduces the number of personnel required to be dispatched to the field but also reduces user downtime, creating a win-win situation.

The present invention discloses a preferred embodiment. Any partial changes or modifications that are derived from the technical concept of the present invention and are easily inferred by a person skilled in the art do not deviate from the scope of the patent rights of the present invention.

In summary, this case demonstrates technical features that are distinct from the known in terms of purpose, means, and effect. Furthermore, it is the first invention of its kind and is suitable for practical use, thus meeting the patent requirements for invention. We earnestly request the Review Committee to carefully examine this matter and to grant a patent as soon as possible, so that it can benefit society and be truly appreciated.

100: Meta-learning-based adaptive computer numerical control compensation system for thermal displacement 10: Device communication module 20: Temperature communication module 30: Displacement communication module 40: Signal processing unit 50: Diagnosis module 60: Deployment module 70: Host system communication module 80: Interface display 200: Device 210: Device controller 220: Temperature sensor 230: Displacement sensor 300: Host system 310: Original model module 320: Update module 330: Meta-learning module S100: Meta-learning-based adaptive computer numerical control compensation method for thermal displacement S11-S18, S191, S192: Steps

Figure 1 is a schematic diagram of the structural blocks of the meta-learning-based thermal displacement adaptive computer numerical control compensation system of the present invention. Figure 2 is a schematic diagram of the structural blocks of the upper system for multiple meta-learning-based thermal displacement adaptive computer numerical control compensation systems of the present invention. Figure 3 is a schematic diagram of the process of the meta-learning-based thermal displacement adaptive computer numerical control compensation method of the present invention. Figure 4 is a schematic diagram of data splicing of the meta-learning-based thermal displacement adaptive computer numerical control compensation system of the present invention. Figure 5 is a schematic diagram of the thermal compensation model verification of the meta-learning-based thermal displacement adaptive computer numerical control compensation system of the present invention.

100: Thermal displacement adaptive computer numerical control compensation system based on meta-learning

10: Device communication module

20: Temperature communication module

30: Displacement Communication Module

40: Signal processing unit

50: Diagnostic Module

60: Deployment Module

70: Host system communication module

80: Interface Display

200: Equipment

210: Device Controller

220: Temperature sensor

230: Displacement Sensor

300: Host system

310: Original Model Module

320: Update module

330: Meta-learning module

Claims (14)

A meta-learning-based thermal displacement adaptive computer numerical control compensation system includes: a device communication module, configured to be electrically connected to a controller of a corresponding device to receive device data of the device and write a compensation value to the device for compensation; a temperature communication module, configured to be electrically connected to at least one temperature sensor located at a predetermined position of the device to receive temperature data read and collected by the at least one temperature sensor; a displacement communication module, configured to be electrically connected to a displacement sensor located on the device to receive displacement data read and collected by the displacement sensor; a signal processing unit, configured to be electrically connected to a controller of a corresponding device to receive device data of the device and write a compensation value to the device for compensation; a temperature communication module, configured to be electrically connected to at least one temperature sensor located at a predetermined position of the device to receive temperature data read and collected by the at least one temperature sensor; a displacement communication module, configured to be electrically connected to a displacement sensor located on the device to receive displacement data read and collected by the displacement sensor; and a signal processing unit, configured to be electrically connected to a controller of a corresponding device to receive device data of the device and write a compensation value to the device for compensation. The device communication module, the temperature communication module, and the displacement communication module are electrically connected to the device communication module, the temperature communication module, and the displacement communication module to process and store the device data, the temperature data, and the displacement data; a diagnosis module is configured to be electrically connected to the signal processing unit and to perform evaluation and diagnosis based on the device data, the temperature data, and the displacement data in a predetermined manner to establish a thermal displacement compensation model; a deployment module is configured to be electrically connected to the signal processing unit and to use a deep learning model to output a prediction value as a compensation benchmark; and a host system communication module is configured to be electrically connected to the signal processing unit and to support a many-to-one communication connection and execute a unary learning host system. The meta-learning-based thermal displacement adaptive computer numerical control compensation system as described in claim 1 further includes an interface display that displays the temperature data, the compensation value, and the displacement data in a corresponding graph through internal software, and is configured to set a compensation mechanism according to user requirements and upload the temperature data, the compensation value, and the displacement data to the host system. The meta-learning-based thermal displacement adaptive computer numerical control compensation system as described in claim 2, wherein the predetermined method includes at least residual analysis and overfitting. As described in claim 3, the meta-learning-based thermal displacement adaptive computer numerical control compensation system, wherein the equipment data at least includes the operating status and processing time of the equipment itself. The meta-learning-based thermal displacement adaptive computer numerical control compensation system as described in claim 4, wherein the predetermined location is inside the device, outside the device, or the location of a key component of the device. As described in claim 5, the meta-learning-based thermal displacement adaptive computer numerical control compensation system, wherein the signal processing unit is a microprocessor having a physical transmission interface of one or more of an electrical potential transmission function, Ethernet or Wi-Fi, and input/output (IO) communication. The meta-learning-based thermal displacement adaptive computer numerical control compensation system as described in claim 1, wherein the host system is a cloud server. As described in claim 7, the meta-learning-based thermal displacement adaptive computer numerical control compensation system includes an original model module, which performs feature engineering on the temperature data and then establishes a basic compensation model. As described in claim 8, the meta-learning-based thermal displacement adaptive computer numerical control compensation system, wherein the upper system further includes an update module that provides the basic compensation model with the meta-learning and data splicing for updating, and then returns the updated basic compensation model to the update module through the deployment module, providing a data analysis and model training function that is relatively unfamiliar to the client. As described in claim 9, the thermal displacement adaptive computer numerical control compensation system based on meta-learning, wherein the upper system further includes a unary learning module, the update module updates the basic compensation model by splicing the meta-learning and the data through the meta-learning module, and the meta-learning module uses a few-sample learning method to learn new tasks using a small amount of data. A meta-learning-based thermal displacement adaptive computer numerical control compensation method includes: collecting temperature data and displacement data; storing the temperature data and the displacement data and uploading them to a cloud server; performing feature analysis on the temperature data and the displacement data; establishing a basic compensation model; and performing thermal error compensation on a device and returning to the step of collecting the temperature data and the displacement data. The step of establishing the basic compensation model simultaneously includes remotely updating the basic compensation model and performing unary learning and data splicing to update it. As described in claim 11, the meta-learning-based thermal displacement adaptive computer numerical control compensation method further includes installing a temperature sensor at a predetermined location of the device and confirming the communication of the temperature sensor before collecting the temperature data and the displacement data. The predetermined location is inside the device, outside the device, or at the location of a key component of the device. The meta-learning-based thermal displacement adaptive computer numerical control compensation method as described in claim 12, wherein, before the step of collecting the temperature data and the displacement data and after installing the temperature sensor at the predetermined location of the device and confirming the communication of the temperature sensor, it also includes installing a displacement sensor on the device and confirming the communication of the displacement sensor. The meta-learning-based thermal displacement adaptive computer numerical control compensation method as described in claim 13, wherein, before the step of collecting the temperature data and the displacement data and after installing the displacement sensor on the device and confirming the communication of the displacement sensor, it also includes using a standard rod for measurement.