TW201818169A - Method for constructing a processing expert system and electronic device using the method capable of obtaining demand indicators like speed index, geometric accuracy index, and surface quality index - Google Patents

Abstract

This invention relates to a method for constructing a processing expert system. The processing expert system is applicable to a CNC machine tool. The method for constructing a processing expert system comprises the following steps: obtaining multiple sets of actual operation samples, each actual operation sample containing existing control parameters and corresponding operational characteristics. The CNC machine tool will derive the results of machining characteristics based on the existing control parameters; the machining characteristics include a machining time, a contour error, and a tracking error. According to the actual operation sample and a machine learning algorithm, an objective function is obtained, so that a processing expert system capable of accurately predicting the processing parameters of the controller based on the processing demand indicators is efficiently constructed, and the processing demand indicators are a speed index, an accuracy index, and a surface quality index.

Description

Method for constructing processing expert system and electronic device using the method

The invention relates to an expert system, in particular to a method for constructing a processing expert system and an electronic device using the method.

Due to different product characteristics, CNC machine tools usually use cutting speed (speed), geometric accuracy (surface accuracy), and surface quality (surface quality) as processing requirements. There are many factors that affect these performance indicators. Among them, the processing parameters (such as maximum acceleration, jerk, and angular speed) used by the CNC controller during operation have a lot of influence on the processing results, so the CNC controller plays a role Whether the processing result can meet the important role of the processing demand target. The control parameters of traditional CNC controllers are usually selected from the factory default settings. Users rarely further consider the machine's assembly and manufacturing, mechanical characteristics, and processing goals to dynamically adjust the controller's processing parameters. With the evolution of society and technology, today's processed products have gradually developed towards a small number, diversified styles, and exquisite workmanship. Customers also have many customized requirements. The processing parameters of a single controller applied with standard settings usually cannot meet the pre-defined requirements. Set processing demand targets. Although the user can adjust the controller processing parameters according to experience or using the data provided by the manufacturer, the user may need to repeatedly test repeatedly and the manufacturer's data is very small. In the end, it not only causes the user's distress and inconvenience, but also may affect the user. Processing efficiency and quality. It can be known from this that how to construct a processing expert system suitable for CNC controllers to provide better controller processing parameters to achieve the desired processing requirements is one of the goals of the relevant industry.

In view of the above-mentioned shortcomings, an object of the present invention is to provide a method for constructing a processing expert system and an electronic device using the method, which can efficiently obtain the correspondence between the processing parameters of the controller and the processing characteristics.

In order to achieve the above object, the present invention provides a method for constructing a processing expert system, which is applicable to a CNC machine tool, and the method for constructing a processing expert system includes the following steps: obtaining a plurality of sets of actual operation samples, each of which The actual operation sample includes multiple existing control parameters and corresponding multiple processing characteristics, and the CNC machine tool is based on the existing control parameters to obtain results with the processing characteristics, and the processing characteristics include a processing time A contour error and a tracking error; based on the actual operating samples and a machine learning algorithm to obtain an objective function to construct the processing expert system, the objective function can be used to obtain the CNC tool based on three processing demand indicators Machine's multiple controller processing parameters, and the three processing demand indicators are a speed indicator, a precision indicator and a surface quality indicator.

In addition, the present invention provides an electronic device suitable for constructing a processing expert system. The processing expert system can be applied to a CNC machine tool. The electronic device includes an input unit and a processing unit coupled to the input unit. The input unit is used to obtain a plurality of sets of actual operation samples, each of which includes a plurality of existing control parameters and corresponding processing characteristics, and the CNC machine tool is based on the existing control parameters. As a result of the processing characteristics, the processing characteristics include a processing time, a contour error, and a tracking error; the processing unit is coupled to the input unit, and the processing unit is based on the actual operation samples and through a machine learning algorithm An objective function can be obtained to construct the processing expert system. The objective function can obtain the processing parameters of multiple controllers of the CNC machine tool according to the three processing demand indicators, and the three processing demand indicators are a speed indicator, a Precision index and a surface quality index.

With the machine learning algorithm, the present invention can automatically and efficiently analyze the correspondence between the processing parameters of the controller and the processing characteristics from the actual operating samples, thereby constructing an accurate prediction based on the three processing demand indicators. Processing expert system of controller processing parameters.

Please refer to FIG. 1. FIG. 1 shows an electronic device 1 according to a preferred embodiment of the present invention. The electronic device 1 includes a display unit 10, an input unit 30, a processing unit 50 coupled to the display unit 10 and the input unit 30, and a storage unit. 70 is coupled to the processing unit 50. The electronic device 1 can be a personal computer, a notebook computer, a smart phone, a tablet computer, a server, or other devices, or it can be built into a CNC machine tool (not shown).

The display unit 10 may be a video display unit such as an LCD, an LED, or the like. The display unit 10 is also used to display a user interface for a user to operate. The user interface is to be described in detail in subsequent embodiments.

The input unit 30 can be any type of unit (such as a mobile communication unit, an Ethernet module, a Bluetooth module, a bus port, etc.) that can receive data through a transmission medium such as a wireless signal, optical fiber, cable, or a bus. ). The input unit 30 may also be combined with a capacitive touch panel, a resistive touch panel, a button, a keyboard, or a mouse to receive input information obtained by a user for input operations on the user interface.

The processing unit 50 may be a central processing unit, a chip, or other programmable general-purpose or special-purpose microcontrollers that have similar functions or a combination of the above-mentioned functions with signal processing and parameter calculation functions. The processing unit 50 is used to: Controls all functions of the electronic device 1.

The storage unit 70 may be a storage medium such as a random access memory, a flash memory, a read-only memory, an erasable programmable read-only memory or an electronic erasable rewritable read-only memory. The unit 70 is used to store various control parameters (such as the maximum acceleration, jerk, angular speed and other CNC controllers (not shown)) of the CNC machine tool, processing parameters, linear, square, and circular machining trajectories, Parameters and data such as the machining path (ie coordinate position information) obtained through the optical ruler (not shown) of the CNC machine tool, the input information, and multiple sets of actual operation samples. Each actual operation sample includes multiple existing control parameters and corresponding multiple processing characteristics (for example, a processing time, a contour error, a tracking error, etc.), and the CNC controller of the CNC machine tool is based on each The controller has the processing parameters of the existing control parameters, and the results have corresponding processing characteristics after actual operation.

The foregoing description is related to the structure and component design of the electronic device 1. The method of constructing a processing expert system for the electronic device 1 will be further described below with reference to FIG. 2. This processing expert system is applicable to the CNC machine tool.

Referring to FIG. 2, in step S21, the processing unit 50 obtains multiple sets of actual operation samples through the input unit 30. For example, the processing unit 50 uses the input unit 30 to obtain the actual operation samples from the Internet, a storage device, other electronic devices, or a user to input data. It should be noted that there may be many implementation examples of the data received by the input unit 30, as long as the processing unit 50 can obtain the actual operation samples.

After the processing unit 50 obtains the actual operation samples under the actual operation of the CNC controller, it can further analyze the correspondence between the processing characteristics and the existing control parameters to learn the changes in the processing parameters of the controller. For each of these processing characteristics, the influence trend. A system capable of understanding the aforementioned correspondence and influence trends is a processing expert system suitable for the CNC controller. In the present invention, the processing unit 50 obtains an objective function based on the actual operation samples and a machine learning algorithm to construct the processing expert system (step S23). The objective function loaded by the processing expert system can obtain the processing parameters of multiple controllers of the CNC machine tool according to the three processing demand indicators. The three processing demand indexes are a speed index, a precision index, and a surface quality index related to the processing efficiency index.

It is worth noting that the machine learning algorithm of this embodiment uses an inverted transfer neural network algorithm. The backward transitive neural network algorithm is an algorithm developed after imitating the biological neural network architecture, which is based on a supervised learning. The processing unit 50 obtains all connection weights between the processing characteristics (output layer neurons) and the existing control parameters (input layer neurons) through the supervised learning and according to the actual operating samples. Including hidden layer neurons), the processing unit 50 constructs the objective function based on the connection weights. For the foregoing construction method, please refer to FIG. 3. The processing unit 50 initializes the connection weights, and obtains multiple estimated characteristics based on the existing control parameters and the connection weights (for example, an estimated processing time, An estimated contour error, an estimated tracking error, etc.). The processing unit 50 calculates an estimation error between the processing characteristics and the estimated characteristics, and adjusts the connection weights by minimizing the estimation error of the actual operation samples. Finally, the processing unit 50 can obtain The objective function. Compared with the use of empirical rules or manufacturer data, the present invention can more efficiently analyze the processing characteristics of the controller processing parameters under different conditions through the machine learning algorithm.

It should be noted that there are actually many kinds of machine learning algorithms. Algorithms such as support vector machines, decision trees, and linear regression can be applied to construct the objective function or form an objective model. The content needs to be adjusted according to different algorithms.

After a processing expert system having the objective function is constructed, the present invention further provides a user interface to facilitate a user to operate the processing expert system loaded in the electronic device 1. In this embodiment, the processing unit 50 displays a user interface with the three processing demand indicators through the display unit 10. Because the three processing demand indicators conflict with each other, the actual processing results cannot make the three processing demand indicators all reach the highest preferred demand at the same time. For example, FIG. 4 is a user interface 80 of a preferred embodiment. The user interface 80 has a triangular area 81 with the speed index, the accuracy index, and the surface quality index as vertices. The user can select a target position P for the three processing demand indicators in the triangular area 81, and the target position P represents the user's preferred demand for each of the processing demand indicators.

It should be noted that the definition of the triangular area 81 is due to the corresponding relationship between the three processing demand indicators. However, there may be many changes in the user interface 80. It may be an interface that can adjust the percentage of the preferred demand. , The option to directly enter the interface of these preferred demand, etc.

When the user selects the target position P in the triangular area 81, the processing unit 50 may receive an input information corresponding to the three processing demand indicators on the user interface 80 through the input unit 30. The input information is Is the target position P in the triangular area 81.

Then, the processing unit 50 calculates the result of the objective function through an optimization algorithm according to the input information to obtain the controller processing parameters corresponding to the input information. In this embodiment, the objective function of the optimization algorithm is defined as equation (1): …(1) Is the objective function, r is the distance between the objective position P and the three vertex positions ( , , ) (I.e. , I is 1, 2, 3), Is a function of processing time (average of processing time for straight, square and circular processing trajectories), Is a contour error function (average of the contour errors of square and circular machining trajectories), Is a tracking error function (average of tracking errors for straight, square, and circular machining trajectories).

The processing unit 50 calculates the inverse of the distance between the target position P and the three vertex positions ( , , ) As the weight value of the three processing demand indicators (that is, the preferred demand volume). The optimization algorithm of this embodiment uses a particle swarm optimization algorithm. The processing unit 50 calculates the minimum value of the objective function (and Limiting the range of the machining parameters of the CNC controller) to obtain the machining parameters of the controller (for example, the machining parameters of the CNC controller). For example, the processing unit 50 defines a solution space according to the objective function, initializes the positions (ie solutions or results) and velocities of a plurality of particles in the solution space, and evaluates the adaptive values of the particles to determine whether to update the particles. The best position of the particles and the best position of the group of the particles is to move the particles, and iteratively search to find the best position of the particles as the control parameters.

It should be noted that the particle swarm optimization algorithm is an algorithm developed using swarm intelligence, which can obtain the best solution in the optimization problem and effectively save costs and time. However, there are actually many kinds of optimization algorithms. In other embodiments, the optimization algorithms can also be changed to simulate degradation, genes and other types of algorithms according to requirements. The objective function and the weight values Need to adjust appropriately to match the corresponding algorithm.

The processing unit 50 can display the processing parameters of the controller in a message display field 83 of the user interface 80 through the display unit 10, or through a communication unit (not shown, for example, using USB, WiFi, etc. Or wireless communication technology) to transmit and set the processing parameters of these controllers to the CNC controller of the CNC machine tool.

And because the different positions in the triangular area 81 correspond to the different preferred demand quantities of the three processing demand indicators, there are many combinations of preferred demand quantities that can be selected by the user, and it is difficult to infer even by the user through rules of thumb or manufacturer data Accurately correspond to the controller processing parameters (that is, the three processing demand indicators are too different from the processing characteristics generated based on the controller processing parameters). In contrast, the embodiment of the present invention combines the machine learning algorithm and the optimization algorithm, which can calculate the processing parameters of different controllers and the three processing demand indicators in a specific numerical range based on some known actual operating samples. The corresponding link (ie, the objective function), and then quickly calculate the best result of the objective function as the corresponding controller processing parameters. Thereby, as long as the user easily selects the target position P in the triangular area 81 according to the desired processing demand index, the processing unit 50 can quickly obtain accurate controller processing parameters (that is, the three processing demand indexes are close to Based on the processing characteristics generated by the processing parameters of the controllers), the performance, processing efficiency and processing quality of the CNC controller can be further improved.

The above is only a description of the embodiments of the present invention, and cannot be used to limit the patent scope of the present invention. For example, any simple structural decoration or change that does not exceed the spirit of the present invention should still fall within the scope of the patent scope of the present invention.

1‧‧‧Processing Expert System

10‧‧‧Display unit

30‧‧‧ input unit

50‧‧‧ processing unit

70‧‧‧Storage unit

S21 ～ S23‧‧‧‧Steps

80‧‧‧ user interface

81‧‧‧ triangle area

83‧‧‧ Message display field

P‧‧‧ target position

‧‧‧distance

FIG. 1 is a component block diagram of a processing expert system according to a preferred embodiment of the present invention. FIG. 2 is a flowchart of a method for constructing a processing expert system according to a preferred embodiment of the present invention. FIG. 3 is a schematic diagram of a learning architecture of a reverse transfer neural network according to a preferred embodiment of the present invention. FIG. 4 is a schematic diagram of a user interface of a preferred embodiment of the present invention.

Claims (12)

A method for constructing a processing expert system is applicable to a CNC machine tool. The method for constructing a processing expert system includes the following steps: obtaining a plurality of sets of actual operation samples, each of which includes a plurality of existing operation samples; There are control parameters and corresponding processing characteristics, and the CNC machine tool obtains the results with the processing characteristics based on the existing control parameters. The processing characteristics include a processing time, a contour error, and a tracking error. ; And based on the actual operating samples and a machine learning algorithm to obtain an objective function to construct the processing expert system, wherein the objective function can be obtained by the CNC machine based on the three processing demand indicators of multiple controller processing Parameters, and the three processing demand indicators are a speed indicator, a precision indicator, and a surface quality indicator. The method for constructing a processing expert system as described in claim 1, wherein the machine learning algorithm is an inverted transfer neural network algorithm, and the steps of generating the objective function according to the actual operation samples and using the machine learning algorithm include : Obtaining multiple estimated characteristics based on the existing control parameters; calculating the estimated errors between the processing characteristics and the estimated characteristics; and minimizing the estimated errors of the actual operating samples to obtain the Objective function. The method for constructing a processing expert system as described in claim 1, after generating the objective function according to the actual operation samples and the machine learning algorithm, the method further includes the following steps: displaying a user interface with the three processing demand indicators Receiving input information corresponding to the three processing demand indicators on the user interface; and calculating the result of the objective function through an optimization algorithm based on the input information to obtain the corresponding numbers of the input information Controller processing parameters. The method for constructing a processing expert system as described in claim 3, wherein the optimization algorithm is a particle swarm optimization algorithm, and the result of the objective function is calculated based on the input information and the optimization algorithm. The steps include: obtaining weight values of the three processing demand indicators according to the input information; and obtaining the controller processing parameters through the particle swarm optimization algorithm according to the weight values. The method for constructing a processing expert system as described in claim 4, wherein the user mask has a triangle region with the speed index, the accuracy index, and the surface quality index as vertices, and the input information is a target in the triangle region. Position, and determining the weight value of the three processing demand index according to the input information includes: obtaining the weight value of the three processing demand index according to the distance between the target position and the three vertex position. The method for constructing a processing expert system as described in claim 5, wherein the weight value of the three processing demand indicators is the inverse of the distance between the target position and the three vertex positions. An electronic device is suitable for constructing a processing expert system. The processing expert system can be applied to a CNC machine tool, and the electronic device includes: an input unit for obtaining multiple sets of actual operation samples, each of which is actually operated The sample contains multiple existing control parameters and corresponding processing characteristics, and the CNC machine tool is based on the existing control parameters to obtain results with the processing characteristics. The processing characteristics include a processing time, a Contour error and a tracking error; and a processing unit coupled to the input unit, the processing unit obtains an objective function based on the actual operating samples and a machine learning algorithm to construct the processing expert system, wherein the objective The function can obtain the processing parameters of multiple controllers of the CNC machine tool according to the three processing demand indexes, and the three processing demand indexes are a speed index, an accuracy index, and a surface quality index. The electronic device according to claim 7, wherein the machine learning algorithm is a backward transfer neural network algorithm, and the processing unit obtains a plurality of estimated characteristics based on the existing control parameters, and the processing unit calculates the The processing unit and the estimated error between the estimated characteristics, the processing unit minimizes the estimated error of the actual operating samples to obtain the objective function. The electronic device according to claim 7, further comprising: a display unit coupled to the processing unit for displaying a user interface having the three processing demand indicators, and the input unit receives a corresponding response on the user interface. Based on the input information of the three processing demand indicators, the processing unit calculates the result of the objective function through an optimization algorithm based on the input information to obtain the controller processing parameters corresponding to the input information. The electronic device according to claim 9, wherein the optimization algorithm is a particle swarm optimization algorithm, and the processing unit obtains a weight value of the three processing demand indicators according to the input information, and the processing unit and According to the weight values, the controller processing parameters are obtained through the particle swarm optimization algorithm. The electronic device according to claim 10, wherein the user mask has a triangle region with the speed index, the accuracy index, and the surface quality index as vertices, and the input information is a target position in the triangle region, and The processing unit obtains a weight value of the three processing demand indicators according to the distance between the target position and the three vertex positions. The electronic device according to claim 11, wherein the weight value of the three processing demand indicators is the inverse of the distance between the target position and the three vertex positions.