TWI860071B - Parameter optimization system for additive manufacturing of product and operational method thereof - Google Patents

Abstract

A parameter optimization system for additive manufacturing of a product and an operational method thereof are provided. The operational method comprises a parameter selection step, a magnetic measurement step, a tensile measurement step, a building step, and an execution step. By using a magnetic analyzer and a tensile testing machine to conduct multiple effective experiments to establish a machine learning model, the efficiency of metal laminate manufacturing can be improved.

Description

Parameter optimization system for manufacturing a product by layering and operation method thereof

The present invention relates to a parameter optimization system and an operating method thereof, and in particular to a parameter optimization system and an operating method thereof for layered manufacturing of a product.

Metal additive manufacturing has been introduced into industry in recent years. Metal additive manufacturing is a process in which a concentrated flow of metal powder or wire is locally melted onto a substrate using a high-power laser. In this way, material can be added to the underlying part. The method is suitable for controlled material structuring, and the laser-generated parts are characterized by a dense microstructure that is usually free of pores.

With the improvement of the performance of electrical/electronic parts and the increase of insulating soft magnetic metal powder for forming motor cores, toroidal cores, etc. through lamination manufacturing, it has been required to reduce iron loss and increase magnetic permeability. In order to enhance the magnetic permeability, it is required to reduce the thickness of the insulating layer so as to narrow the space between the soft magnetic metal powder particles. Iron loss is generally composed of hysteresis loss and eddy current loss, and hysteresis loss varies with the type of soft magnetic material, impurity concentration, processing stress, etc. Eddy current loss varies with the resistivity of the soft magnetic material and the completeness of the insulating film. However, reducing the thickness of the insulating layer will affect the mechanical strength of electronic components.

Therefore, in order to overcome the shortcomings and deficiencies in the prior art, the present invention needs to provide an improved parameter optimization system and its operation method for laminated manufacturing of a product to solve the problems existing in the above-mentioned conventional technology.

The main purpose of the present invention is to provide a parameter optimization system and its operation method for lamination manufacturing of a product, using a magnetic analyzer and a tensile testing machine to conduct multiple sets of effective experiments to establish a machine learning model, which can improve the efficiency of metal lamination manufacturing.

To achieve the above-mentioned purpose, the present invention provides an operating method of a parameter optimization system for lamination manufacturing a product, the operating method comprising a parameter selection step, a magnetic property measurement step, a mechanical stretching measurement step, a model building step and a model execution step. In the parameter selection step, a selection module is used to select a plurality of controllable parameters, the controllable parameters including melting chamber oxygen concentration, laser power, laser scanning speed and scanning line spacing; in the magnetic property measurement step, a magnetic analyzer is used to conduct experiments on a plurality of first test pieces according to the controllable parameters to obtain a magnetic property data. material, wherein the magnetic property data includes a magnetic permeability and an iron loss; in the mechanical tensile measurement step, a tensile testing machine is used to conduct experiments on multiple second test pieces according to the controllable parameters to obtain a tensile property data, wherein the tensile property data includes a maximum tensile stress; in the model establishment step, a machine learning module is used to process the tensile property data and the magnetic property data to generate a training model; in the model execution step, the controllable parameters are input into the training model, and the training model processes and outputs multiple property estimation values of the product.

In one embodiment of the present invention, in the parameter selection step, the selection module adjusts the controllable parameters through a Taguchi method and arranges the controllable parameters in a straight line table.

In one embodiment of the present invention, in the magnetic property measurement step, the first test pieces are tested according to the controllable parameters in the orthogonal table, and each first test piece includes an annular body, an excitation winding, and a measuring winding. The excitation winding and the measuring winding are fixed on the annular body and face each other.

In one embodiment of the present invention, in the magnetic property measurement step, the first test pieces are tested at multiple frequencies, including 50 Hz, 200 Hz, 400 Hz and 800 Hz.

In one embodiment of the present invention, in the mechanical tensile measurement step, the second test pieces are tested according to the controllable parameters in the orthogonal table, and each second test piece includes a middle section and two fixed sections, the fixed sections are located on two opposite sides of the middle section, and a transverse cross-sectional area of the fixed sections is larger than a transverse cross-sectional area of the middle section.

In one embodiment of the present invention, the tensile testing machine stretches the fixed sections of the second test piece at a constant speed of 2 μm/s.

In one embodiment of the present invention, in the model building step, the machine learning module introduces a random search method to find the best hyperparameter, and builds the training model based on the hyperparameter.

In one embodiment of the present invention, in the model execution step, the training model obtains the best estimated values of the characteristics based on the input controllable parameters and the output estimated values of the characteristics through a non-climbing sorting genetic algorithm.

In one embodiment of the present invention, the machine learning module uses a gradient boosting decision tree for calculation.

To achieve the above-mentioned purpose, the present invention provides a parameter optimization system for lamination manufacturing of a product, the parameter optimization system comprising a selection module, a magnetic analyzer, a tensile testing machine and a machine learning module; the selection module is configured to select a plurality of controllable parameters, the controllable parameters including melting chamber oxygen concentration, laser power, laser scanning speed and scanning line spacing; the magnetic analyzer is configured to receive the controllable parameters, and conduct experiments on a plurality of first test pieces according to the controllable parameters to obtain a magnetic property data, wherein the magnetic property data includes a magnetic permeability and an iron loss; the tensile testing machine is configured to receive the controllable parameters, and conduct experiments on multiple second test pieces according to the controllable parameters to obtain a tensile property data, wherein the tensile property data includes a maximum tensile stress; the machine learning module is configured to receive the tensile property data and the magnetic property data, and process the tensile property data and the magnetic property data to generate a training model; wherein the controllable parameters are input into the training model, and the training model processes and outputs multiple property estimates of the product.

As described above, the parameter optimization system for lamination manufacturing of a product of the present invention conducts multiple effective experiments through the magnetic analyzer and the tensile testing machine, and discusses the influence of process parameters on product characteristics according to the results of batch experiments, and then establishes a machine learning model to predict material characteristics after accumulating a certain amount of data, thereby inferring product characteristics through the setting of process parameters, thereby improving the efficiency of metal lamination manufacturing printing, reducing redundant experiments, and verifying the accuracy of the training model to achieve the requirements of parameter optimization.

In order to make the above and other purposes, features and advantages of the present invention more clearly understandable, the following will specifically cite the embodiments of the present invention and provide a detailed description in conjunction with the attached drawings. Furthermore, the directional terms mentioned in the present invention, such as top, bottom, top, bottom, front, back, left, right, inside, outside, side, periphery, center, horizontal, transverse, vertical, longitudinal, axial, radial, topmost or bottommost, etc., are only referenced to the directions of the attached drawings. Therefore, the directional terms used are used to explain and understand the present invention, not to limit the present invention.

Please refer to FIG. 1, which is a parameter optimization system for lamination manufacturing of a product according to an embodiment of the present invention, wherein a machine is used for selective laser melting lamination manufacturing to form the product. The parameter optimization system includes a selection module 2, a magnetic analyzer 3, a tensile testing machine 4 and a machine learning module 5. The present invention will describe in detail the detailed structure, assembly relationship and operation principle of each component below.

Continuing with reference to FIG. 1 , the selection module 2 is configured to select a plurality of controllable parameters, wherein the controllable parameters include melting chamber oxygen concentration, laser power, laser scanning speed, and scanning line spacing. In this embodiment, the selection module 2 adjusts the controllable parameters through a Taguchi method and arranges the controllable parameters in a straight line table. For example, as shown in the following table, the parameters A, B, C, and D represent four process parameter factors (such as melting chamber oxygen concentration, laser power, laser scanning speed, and scanning line spacing), so as to conduct multiple effective experiments and discuss the effects of the four process parameters on product characteristics according to the results of batch experiments, and accumulate the data required for machine learning in this process.

Continuing with reference to FIG. 1 , the magnetic analyzer 3 is electrically connected to the selected module 2, and the magnetic analyzer 3 is configured to receive the controllable parameters and perform experiments on a plurality of first test pieces 6 according to the controllable parameters to obtain a magnetic property data, wherein the magnetic property data includes a magnetic permeability and an iron loss.

As shown in FIG. 2 , in this embodiment, since the magnetic properties of soft magnetic materials are related to the operating frequency, the characteristics of products made of soft magnetic materials have different operating frequencies in different application scenarios. The present invention selects four frequencies (50 Hz, 200 Hz, 400 Hz, and 800 Hz), that is, the first test pieces 6 are tested at multiple frequencies, wherein the frequencies are 50 Hz, 200 Hz, 400 Hz, and 800 Hz, and the first test pieces 6 are tested according to the controllable parameters in the orthogonal table. Specifically, each first test piece 6 includes an annular body 61, an excitation winding 62, and a measuring winding 63, and the excitation winding 62 and the measuring winding 63 are fixed on the annular body 61 and are opposite to each other. For example, the excitation winding 62 is set as a 20-turn primary winding, and the measuring winding 63 is set as a 5-turn secondary winding.

As shown in FIG. 1 , the tensile testing machine 4 is electrically connected to the selected module 2, and the tensile testing machine 4 is configured to receive the controllable parameters and perform experiments on a plurality of second test pieces 7 according to the controllable parameters to obtain a tensile property data, wherein the tensile property data includes a maximum tensile stress.

As shown in FIG. 3 , in this embodiment, the second test pieces 7 are tested according to the controllable parameters in the orthogonal table, wherein each second test piece 7 comprises a middle section 71 and two fixed sections 72, the fixed sections 72 are located at two opposite sides of the middle section 71, and a transverse cross-sectional area of the fixed sections 72 is larger than a transverse cross-sectional area of the middle section 71. For example, the tensile testing machine 4 stretches the fixed sections 72 of the second test piece 7 in the opposite direction at a constant speed of 2 μm/s.

As shown in FIG1 , the machine learning module 5 is electrically connected to the magnetic analyzer 3 and the tensile testing machine 4, and the machine learning module 5 is configured to receive the tensile characteristic data and the magnetic characteristic data, and process the tensile characteristic data and the magnetic characteristic data to generate a training model. In this embodiment, the machine learning module 5 can be calculated by the KNN (Nearest Neighbors) algorithm, or by using a gradient boosting decision tree, such as SVM (support vector machine), XGBoost (eXtreme Gradient Boosting), CatBoost (categorical boosting) and LightGBM (Light Gradient Boosting Machine), and the machine learning module 5 introduces a random search method to find the best hyperparameter, and establishes the training model based on the hyperparameter.

In this embodiment, the controllable parameters are input into the training model through the user interface 101 (see FIG. 5 ), and the training model processes and outputs multiple property estimates of the product, and the training model obtains the best property estimates based on the input controllable parameters and the output property estimates through a non-overlapping genetic algorithm. In other words, the training model can obtain an estimate of the product characteristics based on the input parameters, and can infer the product characteristics through the setting of process parameters before the actual printing of the product.

As described above, the parameter optimization system for lamination manufacturing of a product of the present invention conducts multiple effective experiments through the magnetic analyzer 3 and the tensile testing machine 4, and discusses the influence of the process parameters on the product characteristics according to the results of the batch experiments, and then establishes a machine learning model to predict the material characteristics after accumulating a certain amount of data, thereby inferring the product characteristics through the setting of process parameters, thereby improving the efficiency of metal lamination manufacturing printing, reducing redundant experiments, and verifying the accuracy of the training model, so as to achieve the requirements of parameter optimization and obtain product characteristics of low iron loss, high magnetic permeability and high mechanical strength.

Please refer to FIG. 4 and FIG. 1 for an operation method of a parameter optimization system for lamination manufacturing a product according to an embodiment of the present invention. The operation method is operated according to the parameter optimization system, wherein the operation method includes a parameter selection step S201, a magnetic property measurement step S202, a mechanical stretching measurement step S203, a model establishment step S204 and a model execution step S205. The present invention will explain in detail the relationship between each step and its operating principle below.

Continuing to refer to FIG. 4 and FIG. 1 , in the parameter selection step S201, a selection module 2 is used to select multiple controllable parameters, and the controllable parameters include melting chamber oxygen concentration, laser power, laser scanning speed, and scanning line spacing. In this embodiment, the selection module 2 adjusts the controllable parameters through a Taguchi method and arranges the controllable parameters in a straight table.

Continuing to refer to FIG. 4 and FIG. 1 , in the magnetic property measurement step S202, a magnetic analyzer 3 is used to perform experiments on a plurality of first test pieces 6 according to the controllable parameters to obtain magnetic property data, wherein the magnetic property data includes a magnetic permeability and an iron loss. In this embodiment, since the magnetic properties of soft magnetic materials are related to the operating frequency, the characteristics of products made of soft magnetic materials have different operating frequencies in different application scenarios. The present invention selects four frequencies, 50 Hz, 200 Hz, 400 Hz and 800 Hz, that is, the first test pieces 6 are tested at multiple frequencies, wherein the frequencies are 50 Hz, 200 Hz, 400 Hz and 800 Hz, and the first test pieces 6 are tested according to the controllable parameters in the orthogonal table. Each first test piece 6 includes an annular body 61, an excitation winding 62, and a measuring winding 63. The excitation winding 62 and the measuring winding 63 are fixed on the annular body 61 and are opposite to each other. For example, the excitation winding 62 is set as a 20-turn primary winding, and the measuring winding 63 is set as a 5-turn secondary winding.

Continuing to refer to FIG. 4 and FIG. 1 , in the mechanical stretching measurement step S203, a stretching tester 4 is used to test a plurality of second test pieces 7 according to the controllable parameters to obtain a stretching characteristic data, wherein the stretching characteristic data includes a maximum stretching stress. In this embodiment, the second test pieces 7 are tested according to the controllable parameters in the orthogonal table, wherein each second test piece 7 includes a middle section 71 and two fixed sections 72, the fixed sections 72 are located on two opposite sides of the middle section 71, and a transverse cross-sectional area of the fixed sections 72 is greater than a transverse cross-sectional area of the middle section 71. For example, the stretching tester 4 stretches the fixed sections 72 of the second test piece 7 in the opposite direction at a constant speed of 2μm/s.

Continuing to refer to FIG. 4 and FIG. 1 , in the model building step S204, a machine learning module 5 is used to process the tensile property data and the magnetic property data to generate a training model. In this embodiment, the machine learning module 5 uses a gradient boosting decision tree for calculation, and the machine learning module 5 introduces a random search method to find the best hyperparameter, and establishes the training model based on the hyperparameter.

Continuing to refer to FIG. 4 and FIG. 1 , in the model execution step S205, the controllable parameters are input into the training model through the user interface 101 (see FIG. 5 ), and the training model processes and outputs multiple characteristic estimation values of the product. In this embodiment, the training model obtains the best characteristic estimation values based on the input controllable parameters and the output characteristic estimation values through a non-overlapping ranking genetic algorithm.

The present invention combines the Non-dominated Sorting Genetic Algorithm (NSGA-II), one of the genetic algorithms (GA), to recommend process parameters by specifying product characteristics. First, based on the previously established model (input process parameters, output product characteristics), the optimized process parameter recommendations are obtained through continuous evolution and mating. The objective function used as the judgment basis is the previously established training model, i.e., the machine learning model.

As described above, the parameter optimization system for lamination manufacturing of a product of the present invention conducts multiple effective experiments through the magnetic analyzer 3 and the tensile testing machine 4, and discusses the influence of the process parameters on the product characteristics according to the results of the batch experiments, and then establishes a machine learning model to predict the material characteristics after accumulating a certain amount of data, thereby inferring the product characteristics through the setting of process parameters, thereby improving the efficiency of metal lamination manufacturing printing, reducing redundant experiments, and verifying the accuracy of the training model, so as to achieve the requirements of parameter optimization and obtain product characteristics of low iron loss, high magnetic permeability and high mechanical strength.

Although some aspects have been described in the context of a system, it should be understood that the aspects also represent a description of a corresponding method, and therefore, a block or structural element of the system should also be understood as a corresponding method step or a feature of a method step. By analogy, aspects that have been described in the context of a method step or as a method step also represent a description of a corresponding block or detail or feature of a corresponding device. Some or all of the method steps can be performed using a hardware device such as a microprocessor, a programmable computer or an electronic circuit. In some embodiments, some or several of the most important method steps can be performed by such a device.

Depending on the specific implementation requirements, the embodiments of the present invention can be implemented in hardware or in software. The implementation can be implemented when using a digital storage medium, such as a floppy disk drive, DVD, Blu-ray disk drive, CD, ROM, PROM, EPROM, EEPROM or FLASH memory, a hard disk or any other magnetic or optical storage device, which has an electronically readable control signal stored therein, which can cooperate with a programmable computer system, or collaborate, so that the corresponding method is executed. This is why the digital storage medium can be computer readable. Therefore, some embodiments of the present invention include a data carrier, which includes an electronically readable control signal that can cooperate with a programmable computer system to execute any method described herein. Generally, embodiments of the present invention can be implemented as a computer program product having a program code, and when the computer program product is run on a computer, the program code can effectively execute any method. For example, the program code can also be stored on a machine-readable carrier. Other embodiments include a computer program for executing any method described herein, which is stored on a machine-readable carrier. In other words, one embodiment of the method of the present invention is a computer program having a program code for executing any method described herein, when the computer program is run on a computer. Therefore, another embodiment of the method of the present invention is a data carrier (or digital storage medium or computer-readable medium) in which a computer program for executing any method described herein is recorded. Data carriers, digital storage media or recording media are generally tangible or non-volatile. Therefore, another embodiment of the method of the invention is a data stream or signal sequence representing a computer program for executing any of the methods described herein. The data stream or signal sequence can be configured to be transmitted, for example, via a data communication link, such as via a network. Further embodiments include a processing unit, such as a computer or a programmable logic device, configured or suitable for executing any of the methods described herein. Further embodiments include a computer on which a computer program for executing any of the methods described herein is installed.

Another embodiment according to the invention includes a device or system configured to transmit a computer program for performing at least one of the methods described herein to a receiver. For example, the transmission may be electrical or optical. For example, the receiver may be a computer, a mobile device, a storage device, or the like. For example, the device or system may include a file server for transmitting the computer program to the receiver. In some embodiments, a programmable logic device (e.g., a field programmable gate array, FPGA) may be used to perform some or all of the functions of the methods described herein. In some embodiments, the field programmable gate array may cooperate with a microprocessor to perform any of the methods described herein. Generally, in some embodiments, the methods are performed by any hardware device. The hardware device may be any general-purpose hardware such as a computer processor (CPU), or method-specific hardware such as an ASIC.

Although the present invention has been disclosed by way of embodiments, it is not intended to limit the present invention. Anyone skilled in the art can make various changes and modifications without departing from the spirit and scope of the present invention. Therefore, the scope of protection of the present invention shall be subject to the scope of the patent application attached hereto.

101: User Interface

2: Select module

3: Magnetic analyzer

4: Tensile testing machine

5: Machine learning module

6: First test piece

61: Ring-shaped body

62: Excitation winding

63: Measurement winding

7: Second test piece

71: Middle section

72: Fixed segment

S201: Parameter selection step

S202: Magnetic property measurement step

S203: Mechanical tensile measurement step

S204: Model building step

S205: Model execution steps

FIG1 is a schematic diagram of a parameter optimization system for layered manufacturing of a product according to an embodiment of the present invention.

FIG2 is a schematic diagram of a first test piece of a parameter optimization system for layered manufacturing of a product according to an embodiment of the present invention.

FIG3 is a schematic diagram of a second test piece of a parameter optimization system for layered manufacturing of a product according to an embodiment of the present invention.

FIG4 is a flow chart of an operation method of a parameter optimization system for layered manufacturing of a product according to an embodiment of the present invention.

FIG5 is a schematic diagram of a user interface of a parameter optimization system for layered manufacturing of a product according to an embodiment of the present invention.

S201: Parameter selection step

S202: Magnetic property measurement step

S203: Mechanical tensile measurement step

S204: Model building step

S205: Model execution steps

Claims (10)

An operating method of a parameter optimization system for lamination manufacturing a product, the operating method comprising: a parameter selection step, using a selection module to select a plurality of controllable parameters, wherein the controllable parameters include melting chamber oxygen concentration, laser power, laser scanning speed, and scanning line spacing; a magnetic property measurement step, using a magnetic analyzer to experiment on a plurality of first test pieces according to the controllable parameters to obtain a magnetic property data, wherein the magnetic property data includes a magnetic permeability and an iron loss; a mechanical tensile measurement step, using a tensile testing machine to experiment on a plurality of second test pieces according to the controllable parameters to obtain a tensile property data, wherein the tensile property data includes a maximum tensile stress; A model building step, using a machine learning module to process the tensile property data and the magnetic property data to generate a training model; and a model execution step, inputting the controllable parameters into the training model, processing the training model and outputting multiple property estimation values of the product. The operating method of the parameter optimization system for layered manufacturing of a product as described in claim 1, wherein in the parameter selection step, the selection module adjusts the controllable parameters through a Taguchi method and arranges the controllable parameters in an orthogonal table. As described in claim 2, the operating method of the parameter optimization system for laminated manufacturing of a product, in the magnetic property measurement step, the first test pieces are experimented according to the controllable parameters in the orthogonal table, and each first test piece includes an annular body, an excitation winding, and a measuring winding, and the excitation winding and the measuring winding are fixed on the annular body and opposite to each other. In the operating method of the parameter optimization system for layered manufacturing of a product as described in claim 3, in the magnetic property measurement step, the first test pieces are tested at multiple frequencies, including 50 Hz, 200 Hz, 400 Hz and 800 Hz. As described in claim 2, the operating method of the parameter optimization system for layered manufacturing of a product, in the mechanical tensile measurement step, the second test pieces are tested according to the controllable parameters in the orthogonal table, each second test piece includes a middle section and two fixed sections, the fixed sections are located on two opposite sides of the middle section, and a transverse cross-sectional area of the fixed sections is larger than a transverse cross-sectional area of the middle section. An operating method for a parameter optimization system for layered manufacturing of a product as described in claim 5, wherein the tensile testing machine stretches the fixed sections of the second test piece at a constant speed of 2 μm/s. As described in claim 1, the operating method of the parameter optimization system for layered manufacturing of a product comprises: in the model building step, the machine learning module introduces a random search method to find the best hyperparameter, and builds the training model based on the hyperparameter. As described in claim 1, the operating method of the parameter optimization system for layered manufacturing of a product comprises the following steps: in the model execution step, the training model obtains the best estimated values of the characteristics based on the input controllable parameters and the output estimated values of the characteristics through a non-climbing sorting genetic algorithm. A method for operating a parameter optimization system for layered manufacturing of a product as described in claim 1, wherein the machine learning module uses a gradient boosting decision tree for calculation. A parameter optimization system for lamination manufacturing a product, the parameter optimization system comprising: A selection module, configured to select a plurality of controllable parameters, wherein the controllable parameters include melting chamber oxygen concentration, laser power, laser scanning speed, and scanning line spacing; A magnetic analyzer, configured to receive the controllable parameters, and to perform experiments on a plurality of first test pieces according to the controllable parameters to obtain a magnetic property data, wherein the magnetic property data includes a magnetic permeability and an iron loss; A tensile testing machine, configured to receive the controllable parameters, and to perform experiments on a plurality of second test pieces according to the controllable parameters to obtain a tensile property data, wherein the tensile property data includes a maximum tensile stress; and A machine learning module is configured to receive the tensile property data and the magnetic property data, and process the tensile property data and the magnetic property data to generate a training model; The controllable parameters are input into the training model, and the training model processes and outputs multiple property estimates of the product.