TWI761258B - Intelligent thermal displacement compensation system and thermal displacement model establishment and compensation method of processing machine - Google Patents

Abstract

An intelligent thermal displacement compensation system and thermal displacement model establishment and compensation method of processing machine are provided. The system includes an input module receiving the measurement information of a temperature sensor and a thermal displacement sensor; a process module establishing an initial thermal error calculation model and a plurality of feature subsets, obtaining an adaptive error value, then comparing the adaptive error value with a tolerance value, and obtaining an optimal feature subset when the adaptive error value is smaller than the tolerance value, or adjusting the feature subset otherwise; an output module establishing an optimal thermal error calculation model based on the optimal feature subset; and a compensation module selecting a temperature sensing point information to input to the optimal thermal error calculation model, estimating the actual thermal error variable to produce a compensation result, and accordingly compensating the thermal displacement of the processing machine.

Description

Intelligent thermal displacement compensation system for processing machine and thermal displacement model establishment and compensation method

The present invention relates to a related technology of thermal displacement compensation, in particular to an intelligent thermal displacement compensation system of a processing machine and a method for establishing a thermal displacement model.

In order to pursue high productivity and output of processing machines, machine tool related industries will improve the operating efficiency of processing machines to produce more workpieces within a certain operating time without affecting product quality. However, the thermal energy generated by the processing machine in the process of high-speed operation, the heat transfer, thermal diffusion and thermal division between the processing machine mechanisms, and thermal errors caused by various heat source factors such as ambient temperature, will lead to processing. The metal casting of the machine expands and deforms due to heat, and causes the relative position of the tool and the workpiece to shift.

The deformation and offset of the processing machine due to thermal error will affect the processing accuracy of the overall processing machine by 40~70%. Therefore, the industry usually performs thermal error compensation (Thermal Error Compensation) on the processing machine to reduce the thermal error. The effect of precision. Thermal error compensation technology is divided into active compensation and passive compensation. Active compensation usually refers to the structural improvement of the processing machine, and passive compensation refers to the compensation by computer software control.

Although the active compensation method can solve or suppress the influence of thermal error from the source, because it involves the design, assembly and testing of the processing machine, in addition to the high cost and time-consuming, it is impossible to quickly verify the thermal error compensation effect; passive compensation method By installing multiple sets of temperature sensors and arranging displacement sensors on the machine, and establishing a thermal error model accordingly, the thermal error model can estimate the thermal error variable according to the current temperature change for compensation, and can be confirmed immediately Compensation effect.

However, in order to improve the prediction accuracy of the model, it is necessary to screen out the key temperature measurement positions first, and then enter into the model establishment and training procedures such as model-related parameter adjustment, and the key measurement positions and model-related parameters can only be obtained by personnel The intervention is repeatedly adjusted by try and error until the accuracy reaches the standard. Therefore, in addition to the tedious and labor-intensive modeling process, the reliability of the model is also difficult to measure.

The main purpose of the present invention is to screen out the key temperature measurement positions and adjust the self-correlated parameters of the model in the process of model establishment and training. thermal displacement compensation effect.

In order to achieve the above object, an embodiment of the present invention provides an intelligent thermal displacement compensation system for a processing machine. The processing machine has a plurality of temperature sensors located at different positions and a thermal displacement sensor. The plurality of temperature sensors and the thermal displacement sensors are coupled, and the thermal displacement compensation system includes: an input module, a processing module, an output module and a compensation module; the input module is used for receiving a plurality of temperature sensing The measurement information of the device and the thermal displacement sensor; the processing module is coupled to the input module, the processing module includes a model building unit, a feature subset building unit and a training unit, and the model building unit is used to build a The initial thermal error calculation model of temperature change, the feature subset establishment unit obtains complex feature subsets according to the complex temperature sensor and the initial thermal error calculation model, each feature subset includes complex temperature sensing point information and a related initial The model parameters of the thermal error calculation model, the training unit obtains a simulated thermal error variable for each feature subset from the initial thermal error calculation model, and further obtains an adaptive error value based on the simulated thermal error variable, and the training unit compares each The adaptive error value of the feature subset is used to obtain a better feature subset. When the adaptive error value of the better feature subset is less than a tolerance value, the better feature subset is used as an optimum feature subset. When the adaptive error value of the subset is greater than the tolerance value, the complex feature subset is re-adjusted; the output module is coupled with the processing module, and the output module establishes an optimal thermal error according to the optimal feature subset and the initial thermal error calculation model Calculation model; the compensation module is coupled with the input module and the output module, the compensation module can select the temperature sensing point information and input it into the optimal thermal error calculation model, estimate the actual thermal error variable to generate a compensation result, and Thermal displacement compensation is performed on the processing machine according to the compensation result.

An embodiment of the present invention provides a method for establishing an intelligent thermal displacement model of a processing machine. The processing machine has a plurality of temperature sensors located at different positions and a thermal displacement sensor. The method for establishing the thermal displacement model includes the following steps: A model establishment step, a data acquisition step, a thermal error simulation step, a model training step, and an optimization step; the model establishment step: establishing an initial thermal error calculation model according to temperature changes; the data acquisition step: according to The complex temperature sensors and the initial thermal error calculation model on the processing machine are used to obtain complex feature subsets, each feature subset includes the temperature sensing point information obtained by the temperature sensor and a related initial thermal error calculation Model parameters of the model; thermal error simulation step: the initial thermal error calculation model obtains a simulated thermal error variable according to the complex feature subset, and further obtains an adaptive error value based on the simulated thermal error variable; model training step: comparison The adaptive error value of each feature subset is used to obtain a better feature subset. When the adaptive error value of the better feature subset is less than a tolerance value, the better feature subset is used as an optimal feature subset, and the The optimal feature subset and the initial thermal error calculation model establish an optimal thermal error calculation model. When the adaptive error value of the best feature subset is greater than the tolerance value, the subsequent steps are performed; the optimization step: readjust the complex features parameter selection of the set, and return to the thermal error simulation step, and so on until in the model training step, the adaptive error value of the preferred feature subset is less than the tolerance value.

Thereby, the present invention establishes an initial thermal error calculation model and feature subset according to the temperature change through the processing module, and further obtains an adaptive error value. The processing module compares the adaptive error value and a tolerance value, and when the adaptive error value is used When it is less than the tolerance value, an optimal feature subset is obtained. When the adaptation error value is greater than the tolerance value, the feature subset is re-adjusted. In this way, the most critical temperature sensing points are simultaneously selected and the model is adjusted to its own model parameters, so as to achieve The effect of simplifying the modeling process and improving the reliability of the model.

In order to facilitate the description of the central idea of the present invention expressed in the column of the above-mentioned summary of the invention, specific embodiments are hereby expressed. Various objects in the embodiments are drawn according to proportions, sizes, deformations or displacements suitable for description, rather than the proportions of actual elements, which will be described first.

Please refer to FIG. 1 to FIG. 3 , the present invention provides an intelligent thermal displacement compensation system and a thermal displacement model establishment and compensation method for a processing machine. The processing machine 1 has a plurality of temperature sensors 2 located at different positions and a thermal Displacement sensor 3, wherein each temperature sensor 2 located at a different position on the processing machine 1 represents a temperature sensing point; the thermal displacement compensation system 100 and the plurality of temperature sensors 2 and the thermal displacement sensor 3 coupled.

The thermal displacement compensation system 100 includes: an input module 10 , a processing module 20 , an output module 30 and a compensation module 40 , wherein the input module 10 is used for receiving a plurality of temperature sensors 2 and thermal displacement sensors The measurement information of the measuring device 3; the processing module 20 is coupled to the input module 10, and the processing module 20 includes a model building unit 21, a feature subset building unit 22 and a training unit 23; the output module 30 and processing The module 20 is coupled; the compensation module 40 is coupled to the input module 10 and the output module 30 .

The thermal displacement model establishment and compensation method 200 of this embodiment includes a model establishment step 201 , a data acquisition step 202 , a thermal error simulation step 203 , a model training step 204 , an optimization step 205 and a compensation step 206.

In the model building step 201 , the model building unit 21 builds an initial thermal error calculation model 211 depending on the temperature.

Next, in the data acquisition step 202, the feature subset creation unit 22 obtains a complex feature subset X according to the complex temperature sensing point information and the initial thermal error calculation model 211, and each feature subset X includes a complex temperature sensing Point information and model parameters of an associated initial thermal error calculation model 211 .

In this embodiment, different model parameter information is selected for each feature subset X, and at least one key temperature point information is selected from all the temperature sensing point information, and other temperature sensing point information is excluded from the feature subset. The effect of X.

Among them, the model parameters of this embodiment are the establishment of the initial thermal error calculation model 211, which is based on the support vector regression (SVR, Support Vector Regression) algorithm establishment of the model as an example, the model parameters are to define the kernel function when using the SVR algorithm, Functions or hyperparameters such as C and gamma; SVR projects the feature subset X from 2D to 3D through the kernal function, cuts it through the hyperplane and then maps it back to 2D for classification; By adjusting the parameters C and gamma, the SVR can be adjusted to build a model The allowable error and the SVR fitting effect.

As shown in FIG. 3, it is a schematic diagram of the temperature sensing point selection of the processing machine 1 according to the embodiment of the present invention. When the temperature sensing point is selected as the key temperature point, the temperature sensing point information is displayed in the form of 0 or 1. When a certain temperature sensing point is selected as the key temperature point When the feature of a temperature sensing point is 1, it means that the temperature sensing point is selected as the key temperature point, and the feature subset X uses the key temperature point information of the key temperature point; when the feature is 0, it means that the temperature is not selected The sensing point is a key temperature point, and the feature subset X does not use the temperature sensing point information of this temperature sensing point.

Next, in the thermal error simulation step 203 , the training unit 23 obtains a simulated thermal error variable from the initial thermal error calculation model 211 according to the complex feature subset X, and each feature subset X is further based on the simulated thermal error variable. An adaptive error value 231 is obtained. The training unit 23 compares the simulated thermal error variable with a displacement information measured by the thermal displacement sensor 3 to obtain a test error value, and uses the test error value as the adaptive error value 231 to perform subsequent step processing .

In this embodiment, in the thermal error simulation step 203, the training unit 23 can calculate and obtain the adaptive error value 231 of each feature subset X according to the complex feature subset X, wherein the adaptive error value 231 of each feature subset X is a test. The sum of the error value and the error value of a temperature point.

Further description, the training unit 23 obtains complex analog thermal error variables according to the complex feature subset X, and compares each analog thermal error variable with the displacement information measured by the thermal displacement sensor 3 to calculate and obtain each feature Set the test error value of X, and define the temperature point error value (the number of critical temperature points divided by the number of temperature sensing points).

To further illustrate, in this embodiment, considering the needs of each user for modeling thermal displacement models of different processing machines 1, an adaptation error value 231=w1*test error value+w2*temperature error value is further defined, where w1 is the test error Value weight, w2 is the temperature error value weight, and the sum of weight w1 and weight w2 is 1.

Next, in the model training step 204 , the training unit 23 compares the adaptive error values 231 of each feature subset, and determines a better feature subset according to the magnitude of each adaptive error value 231 . When the adaptive error value 231 is smaller than a tolerance value, it means that the overall thermal displacement simulation result has reached the predetermined target, which is consistent with the actual situation. The optimal feature subset X' and the initial thermal error calculation model 211 establish an optimal thermal error calculation model 31 .

In the model training step 204 , when the adaptation error value 231 of the preferred feature subset is greater than the tolerance value, it means that the overall thermal displacement simulation result has not reached the predetermined target, and the subsequent optimization step 205 will be performed. In the optimization step 205, the training unit 23 re-adjusts the parameter selection of the complex feature subset X, and returns to the thermal error simulation step 203, and repeats the adjustment of the feature subset X in this cycle until in the model training step 204, the more The adaptation error value 231 of the best feature subset is less than the tolerance value until the best feature subset X' is obtained.

Among them, in the optimization step 205 of this embodiment, the method for adjusting the complex feature subset X is to take the Particle Swarm Optimization algorithm as an example, and the complex feature subset X is modified by the velocity vector so that it gradually reaches For optimization, the velocity vector is updated based on the ratio constant term, the random range, the best feature subset and the best feature subset X', and the update method is as follows:

；

;

；

;

為特徵子集X，

為速度向量，

與

為比例常數項，

與

為介於0~1之隨機範圍常數，t為當前迭代次數，

為該次迭代之較佳特徵子集，

為截至該次迭代之當前所有特徵子集X中之最佳特徵子集X’，當

之適應誤差值231小於

之適應誤差值231，即更新成為新的

，而

小於容忍值時即成為最佳特徵子集X’。

is the feature subset X,

is the velocity vector,

and

is the proportional constant term,

and

is a random range constant between 0 and 1, t is the current iteration number,

is the best feature subset for this iteration,

is the best feature subset X' among all feature subsets X up to the current iteration, when

The adaptive error value 231 is less than

The adaptation error value is 231, that is, the update becomes the new

,and

When it is less than the tolerance value, it becomes the best feature subset X'.

Finally, in the compensation step 206, the compensation module 40 can input the critical temperature point information of the critical temperature point to the optimal thermal error calculation model 31 established by the optimal feature subset X' and the initial thermal error calculation model 211. The optimal thermal error calculation model 31 can estimate the actual thermal error variable to generate a compensation result, and perform thermal displacement compensation on the processing machine 1 according to the compensation result.

In this way, the present invention establishes an initial thermal error calculation model 211 according to temperature changes and a complex feature subset X through the processing module 20, and each feature subset X includes complex temperature sensing point information and a related initial thermal error calculation model 211 model parameters. The processing module 20 continuously updates the feature subset X by evaluating the adaptation error value 231 until the best feature subset X' is obtained. The process of updating the feature subset X can simultaneously screen the most critical temperature sensing points and make the model perform self-modeling Parameter adjustment, and without manual intervention, a model with the best thermal displacement compensation effect can be established, thereby achieving the effect of simplifying the modeling process and improving the reliability of the model.

The above-mentioned embodiments are only used to illustrate the present invention, but not to limit the scope of the present invention. All the modifications or changes that do not violate the spirit of the present invention belong to the intended protection category of the present invention.

1: Processing machine

2: temperature sensor

3: Thermal displacement sensor

100: Thermal Displacement Compensation System

10: Input module

20: Processing modules

21: Model building unit

211: Initial Thermal Error Calculation Model

22: Feature subset establishment unit

23: Training Unit

231: adaptation error value

30: Output module

31: Best Thermal Error Calculation Model

40: Compensation module

200: Thermal Displacement Modeling Methods

201: Model building steps

202: Data Capture Steps

203: Thermal Error Simulation Steps

204: Model training steps

205: Optimization Steps

206: Compensation step

X: feature subset

X': best feature subset

FIG. 1 is a block diagram of an intelligent thermal displacement compensation system of a processing machine according to an embodiment of the present invention. FIG. 2 is a flowchart of a method for establishing an intelligent thermal displacement model of a processing machine according to an embodiment of the present invention. FIG. 3 is a schematic diagram of screening of temperature sensing points of a processing machine according to an embodiment of the present invention.

200: Thermal Displacement Modeling Methods

201: Model building steps

202: Data Capture Steps

203: Thermal Error Simulation Steps

204: Model training steps

205: Optimization Steps

206: Compensation step

Claims (13)

An intelligent thermal displacement compensation system for a processing machine, the processing machine has a plurality of temperature sensors located at different positions and a thermal displacement sensor, the thermal displacement compensation system, the plurality of temperature sensors and the thermal displacement sensor The detector is coupled, and the thermal displacement compensation system includes: an input module for receiving the plurality of temperature sensors and measurement information of the thermal displacement sensor; a processing module, which is connected with the input module Coupling, the processing module includes a model building unit, a feature subset building unit and a training unit, the model building unit is used to build an initial thermal error calculation model according to temperature changes, the feature subset building unit is based on each The temperature sensor and the initial thermal error calculation model obtain complex feature subsets, each feature subset includes information of each temperature sensing point and a model parameter related to the initial thermal error calculation model, the training unit is composed of the The initial thermal error calculation model obtains a simulated thermal error variable for each feature subset, and further obtains an adaptive error value based on the simulated thermal error variable, and the training unit compares the adaptive error value of each feature subset to obtain A better feature subset, when the adaptation error value of the better feature subset is less than a tolerance value, the better feature subset is used as an optimum feature subset, when the adaptation error value of the better feature subset is When the error value is greater than the tolerance value, each feature subset is readjusted; an output module is coupled to the processing module, and the output module is established according to the optimal feature subset and the initial thermal error calculation model an optimal thermal error calculation model; and a compensation module coupled to the input module and the output module, the compensation module can select the temperature sensing point information and input it to the optimal thermal error calculation The model estimates the actual thermal error variable to generate a compensation result, and performs thermal displacement compensation on the processing machine according to the compensation result. The intelligent thermal displacement compensation system for a processing machine as claimed in claim 1, wherein each of the In each feature subset, at least one key temperature point information is selected from the temperature sensing point information, and the influence of other temperature sensing point information on the feature subset is excluded. The intelligent thermal displacement compensation system for a processing machine as claimed in claim 1, wherein different model parameter information is selected and used in each of the feature subsets. The intelligent thermal displacement compensation system for a processing machine as claimed in claim 1, wherein the training unit is used for comparing the simulated thermal error variable with a measured displacement information to calculate and obtain a test error value, and The test error value is used as the adaptive error value for subsequent processing. The intelligent thermal displacement compensation system for a processing machine as claimed in claim 1, wherein the training unit compares the simulated thermal error variable with measured displacement information to calculate and obtain a test error value, and define a The temperature point error value is (the number of critical temperature points divided by the number of temperature sensing points), and the adaptive error value is the sum of the test error value and the temperature point error value. The intelligent thermal displacement compensation system for a processing machine as claimed in claim 1, wherein the training unit is used for comparing the simulated thermal error variable with a measured displacement information to calculate and obtain a test error value, and Define a temperature point error value (the number of key temperature points divided by the number of temperature sensing points), and define the adaptive error value = w1*test error value+w2*temperature error value, where w1 is the weight of the test error value, w2 is the weight of the temperature error value, and the sum of the weight w1 and the weight w2 is 1. A method for establishing an intelligent thermal displacement model of a processing machine. The processing machine has a plurality of temperature sensors located at different positions and a thermal displacement sensor. The method for establishing a thermal displacement model comprises the following steps: a model building step: establishing an initial thermal error calculation model according to temperature change; a data acquisition step: obtaining complex feature subsets according to each of the temperature sensors on the processing machine and the initial thermal error calculation model, each feature subset includes The complex number is obtained by the temperature sensor information of each temperature sensing point and a model parameter related to the initial thermal error calculation model; a thermal error simulation step: the initial thermal error calculation model obtains a simulated thermal error variable according to each of the feature subsets, and uses the Based on the simulated thermal error variable, an adaptive error value is further obtained; a model training step: comparing the adaptive error value of each feature subset to obtain a better feature subset, when the adaptive error value of the better feature subset is When it is less than a tolerance value, the better feature subset is used as an optimum feature subset, and an optimum thermal error calculation model is established with the optimum feature subset and the initial thermal error calculation model. When the adaptive error value of the subset is greater than the tolerance value, the following steps are performed; and an optimization step: readjust the parameter selection of each feature subset, and return to the thermal error simulation step, and the cycle is repeated until the In the model training step, the adaptive error value of the preferred feature subset is smaller than the tolerance value. The method for establishing an intelligent thermal displacement model for a processing machine according to claim 7, wherein, in the data acquisition step, at least one key temperature point information is selected from the temperature sensing point information in each of the feature subsets , and exclude the influence of other temperature sensing point information on this feature subset. The method for establishing an intelligent thermal displacement model for a processing machine according to claim 7, wherein, in the data acquisition step, different model parameter information is selected and used in each of the feature subsets. The method for establishing an intelligent thermal displacement model of a processing machine according to claim 7, wherein, in the thermal error simulation step, the simulated thermal error variable is compared with displacement information measured by the thermal displacement sensor , to obtain a test error value by calculation, and use the test error value as the adaptive error value to perform subsequent steps. The method for establishing an intelligent thermal displacement model of a processing machine as claimed in claim 7, wherein, in the thermal error simulation step, a displacement measured by the simulated thermal error variable and the thermal displacement sensor The information is compared to calculate and obtain a test error value, and define a temperature point error value (the number of critical temperature points divided by the number of temperature sensing points), the adaptive error value is the test error value and the temperature point error value Sum. The method for establishing an intelligent thermal displacement model of a processing machine according to claim 7, wherein, in the thermal error simulation step, the simulated thermal error variable is compared with displacement information measured by the thermal displacement sensor , obtain a test error value by calculation, and define a temperature point error value (the number of critical temperature points divided by the number of temperature sensing points), and define the adaptive error value=w1*test error value+w2*temperature error value, Among them, w1 is the weight of the test error value, w2 is the weight of the temperature error value, and the sum of the weight w1 and the weight w2 is 1. A method for thermal displacement compensation using the thermal displacement model of claim 7, further comprising a compensation step: selecting the temperature sensing point information to input into the optimal thermal error calculation model, and estimating the actual thermal error variable to generate a Compensation results, and thermal displacement compensation is performed on the processing machine according to the compensation results.

Images