KR20230172233A - Method for evaluating of joining quality for self piercing rivet based on artificial neural network and system for the same - Google Patents

Abstract

An artificial intelligence learning method that can non-destructively monitor or evaluate the quality of mechanical fastening including self-piercing rivets using artificial intelligence, such as machine learning, an evaluation method of mechanical fastening using the same, and a mechanical fastening facility are provided. The learning method is a method performed by a computing device, measuring acoustic signals generated during the self-piercing riveting process performed in the order of clamping, piercing, and flaring, and performing Fourier transform on at least a portion of the measured acoustic signals. Data based on the obtained converted data is used as learning data.

Description

Artificial neural network-based self-piercing riveting quality evaluation method and system for the same {METHOD FOR EVALUATING OF JOINING QUALITY FOR SELF PIERCING RIVET BASED ON ARTIFICIAL NEURAL NETWORK AND SYSTEM FOR THE SAME}

The present invention relates to a method for non-destructively and real-time monitoring the quality of mechanical fastening including self-piercing rivets, a self-piercing rivet fastening process method including the same, and an equipment system for the same. Specifically, an artificial intelligence learning method that can non-destructively monitor or evaluate the quality of mechanical fastening including self-piercing rivets using artificial intelligence, such as machine learning, a mechanical fastening evaluation method using the same, and a mechanical fastening equipment. is provided.

In general, joining heterogeneous materials or joining multiple materials means joining or combining a plurality of materials with different types, physical properties, and characteristics. Bonding of different materials in a broad sense can be used to include bonding technology between materials, molding technology, surface treatment technology, etc. Recently, as materials are required to be lighter and have higher strength, various materials other than conventional metals are being utilized, and accordingly, various studies related to dissimilar material joining technology are being conducted.

However, existing fusion-based welding technologies for steel materials (e.g., resistance spot welding, arc welding, laser welding, etc.) cannot be applied to joining dissimilar materials with different physical properties, so adhesive bonding is used. Non-welding techniques such as adhesive bonding and mechanical jointing are being proposed.

Mechanical fastening technology is a general term for technologies that maintain physical boundaries between joining materials and mechanically join dissimilar materials using separate fastening members. Traditional mechanical fastening methods, such as bolt-nut coupling and blind riveting, require hole machining of the joining member first, and alignment of holes and holes and rivets is required, so the process is lengthy. It has the disadvantage of being complex and difficult to automate. Therefore, recently, mechanical fastening methods that do not require hole processing, such as self-piercing riveting (SPR), flow drill screwing (FDS), and clinching, have been in the spotlight.

J.P. 2018-010632 A J.P. 2018-054587 A KR 10-2016-0011060 A

Self-piercing riveting (SPR) is a mechanical fastening technology and is one of the representative forming-based spot joining technologies. In the case of self-piercing riveting, it is a widely used mechanical fastening method because it does not cause significant deformation of the material and can provide very strong bonding strength in a relatively simple method.

Self-piercing riveting has the advantage of being highly productive, relatively low in thermal deformation, and applicable to different materials and coated materials because the process is simpler than bolt-nut fastening and can be automated. In addition, it can be used for joining dissimilar materials such as non-ferrous metals and steel, has excellent bonding strength, and is environmentally friendly, so it is evaluated as the most realistic alternative to spot bonding dissimilar materials.

Self-piercing riveting is performed by inserting a specially shaped rivet on one side of two materials to be joined. Self-piercing riveting is where the rivet penetrates the upper material, and the shape of the rivet (called flaring) is deformed according to the shape of the molding table (or die) located below the lower material, and the shape of the lower material is deformed accordingly. The material and the underlying material are strongly interlocked and combined. Therefore, self-piercing riveting may cause fastening defects when the ductility of the lower material is low, the strength of the upper material is too high, or mis-alignment occurs.

Meanwhile, one of the important elements of mechanical fastening technology is the evaluation of mechanical fastening quality. In particular, in the case of self-piercing riveting, etc., once fastening is made, there is no suitable means to evaluate the joint state, that is, the quality of fastening, without destroying it. In particular, from the perspective of automation of mechanical fastening, there is a great need for non-destructive inspection of the fastening area after fastening and further real-time monitoring, and various technologies are being developed accordingly. For example, Patent Document 3 discloses a method for inspecting riveted joints of metal panels. Specifically, Patent Document 3 discloses a method of evaluating fastening quality using the load and displacement of the punch used in riveted joints.

However, the fastening quality inspection method based on load and displacement, such as in Patent Document 3, has various variables, so it is inconvenient to dataize them and proceed with initial settings. Accordingly, one facility can be used for various types of members. There is a problem that makes it difficult to apply it to quality evaluation of mechanical fastening.

Accordingly, the problem to be solved by the present invention is to provide a rivet fastening facility or device system that can monitor or evaluate the quality of mechanical fastening in a new method different from the conventional one. In other words, it provides a device that can predict fastening quality in a non-destructive way during the riveting process.

Another problem to be solved by the present invention is to provide a method for evaluating or monitoring the quality of mechanical fastening such as rivet fastening.

Another problem that the present invention aims to solve is to provide a learning model based on artificial intelligence for evaluating the quality of the above mechanical fastening.

Another problem that the present invention aims to solve is to provide an artificial intelligence learning method for building the above learning model.

The problems of the present invention are not limited to the technical problems mentioned above, and other technical problems not mentioned will be clearly understood by those skilled in the art from the description below.

The artificial intelligence learning method according to an embodiment of the present invention for solving any of the above problems is a method performed by a computing device, and includes sound generated during the self-piercing riveting process performed in the order of clamping, piercing, and flaring. It is an artificial intelligence learning method for predicting mechanical joint quality that measures a signal and uses data based on the converted data obtained by Fourier transforming at least some of the measured acoustic signals as learning data.

The learning data may be an image of converted data with frequency on the horizontal axis and amplitude on the vertical axis.

In some alternative embodiments, the learning data divides the converted data into a plurality of frequency sections, and includes the number of peaks in each section, the value of each peak, the area of each peak, the average value of the peak values in each section, and the number of peaks in each section. One or more of the standard deviation of the peak value within the section and the area of each section may be used.

In addition, the learning data uses the area of at least one peak occurring in the range of at least 100 Hz to 550 Hz, or the area of the range section, and the area of one or more peaks occurring in the range of 2,000 Hz to 2,550 Hz, or the area of the range section. It may be.

In addition, the learning data may use at least the value or number of peaks occurring in the range of 100 Hz to 550 Hz, and the value or number of peaks occurring in the range of 2,000 Hz to 2,550 Hz.

Furthermore, the learning data may use the area of at least one peak occurring in the range of 100 Hz to 150 Hz, or the area of the range section, and the area of one or more peaks occurring in the range of 400 Hz to 550 Hz, or the area of the range section. there is.

The learning data may use at least the value or number of peaks occurring in the range of 100 Hz to 150 Hz, and the value or number of peaks occurring in the range of 400 Hz to 550 Hz.

In addition, the learning data uses the area of at least one peak occurring in the range of at least 400 Hz to 550 Hz, or the area of the range section, and the area of one or more peaks occurring in the range of 550 Hz to 2,000 Hz, or the area of the range section. You can.

A self-piercing riveting device according to an embodiment of the present invention for solving any of the above problems is a self-piercing riveting device that performs clamping, piercing, and flaring in the order, and the size of the acoustic signal over time during the process. An acoustic measuring unit that measures; a Fourier transform unit that performs Fourier transform on the sound signal; and a quality evaluation unit that evaluates process quality by using the Fourier transformed data as input to a learned learning model.

The quality evaluation method of the riveting process according to an embodiment of the present invention to solve the above-mentioned another problem is a quality evaluation method of a self-piercing riveting process performed in the order of clamping, piercing, and flaring, and is provided to a computing device. It is performed by measuring acoustic signals generated during the riveting process, and converting data obtained by Fourier transforming at least some of the measured acoustic signals as input data for the learned evaluation model.

A method for monitoring the quality of mechanical fastening according to an embodiment of the present invention for solving another problem described above is a method performed by a computing device, in a self-piercing riveting process performed in the order of clamping, piercing and flaring. , Measuring the acoustic signal generated in the process and evaluating the quality of the self-piercing riveting process based on the acoustic signal analyzed through Fourier Transform, wherein the step of measuring and analyzing the acoustic signal , comprising the step of separating the sound generated during the clamping, the sound generated during the piercing, and the sound generated during the flaring, and as a result of the Fourier transform of the sound signal during the piercing, the magnitude of the signal generated in the range of 2,000 Hz to 2,550 Hz If the ratio of the sum of the magnitudes of signals occurring in the range of 100Hz to 550Hz to the sum is 10 or more, preferably 12 or more, it is judged to be normal, and if not, it is judged to be defective.

A method for monitoring the quality of mechanical fastening according to another embodiment of the present invention to solve the above problem is a method performed by a computing device, in the self-piercing riveting process performed in the order of clamping, piercing and flaring. , Measuring the acoustic signal generated in the process and evaluating the quality of the self-piercing riveting process based on the acoustic signal analyzed through Fourier Transform, wherein the step of measuring and analyzing the acoustic signal , comprising the step of separating the sound generated during the clamping, the sound generated during the piercing, and the sound generated during the flaring, and as a result of the Fourier transform of the sound signal during the piercing, the size of the signal having a maximum peak in the range of 100 Hz to 550 Hz If the size is more than 3 times, preferably 4 times or more, the size of the signal with the maximum peak in the range of 2,000 Hz to 2,550 Hz, it is judged as normal, and if not, it is judged as defective.

The quality monitoring method of mechanical fastening according to another embodiment of the present invention to solve the above problem is a method performed by a computing device, and involves a self-piercing riveting process consisting of clamping, piercing, and flaring. Measuring the acoustic signal generated in the process and evaluating the quality of the self-piercing riveting process based on the acoustic signal analyzed through Fourier Transform, comprising measuring and analyzing the acoustic signal. It includes the step of separating the sound generated during the clamping, the sound generated during the piercing, and the sound generated during the flaring, and the result of the Fourier transform of the sound signal during the piercing is the sum of the magnitude of the signal generated in the range of 100 Hz to 150 Hz. For , if the ratio of the sum of the magnitudes of signals generated in the range of 400Hz to 550Hz is less than 1, or preferably less than 0.8, it is judged to be normal, and if not, it is judged to be defective.

The quality monitoring method of mechanical fastening according to another embodiment of the present invention to solve the above problem is a method performed by a computing device, and involves a self-piercing riveting process consisting of clamping, piercing, and flaring. Measuring the acoustic signal generated in the process and evaluating the quality of the self-piercing riveting process based on the acoustic signal analyzed through Fourier Transform, comprising measuring and analyzing the acoustic signal. It includes the step of separating the sound generated during the clamping, the sound generated during the piercing, and the sound generated during the flaring, and as a result of the Fourier transform of the sound signal during the piercing, a signal having a maximum peak in the range of 100 Hz to 150 Hz is obtained. If the size is 2.5 times, preferably 2.8 times or more, the size of the signal with the maximum peak in the range of 400Hz to 550Hz, it is judged to be normal, and if not, it is judged to be defective.

The quality monitoring method of mechanical fastening according to another embodiment of the present invention to solve the above problem is a method performed by a computing device, and involves a self-piercing riveting process consisting of clamping, piercing, and flaring. Measuring the acoustic signal generated in the process and evaluating the quality of the self-piercing riveting process based on the acoustic signal analyzed through Fourier Transform, comprising measuring and analyzing the acoustic signal. It includes the step of separating the sound generated during the clamping, the sound generated during the piercing, and the sound generated during the flaring, and the result of the Fourier transform of the sound signal during the piercing is the sum of the magnitude of the signal generated in the range of 400 Hz to 550 Hz. If the ratio of the sum of the sizes of signals generated in the range of 550Hz to 2,000Hz is 0.5 or less, preferably 0.3 or less, it is judged as normal, and if not, it is judged as defective.

Specific details of other embodiments are included in the detailed description.

According to embodiments of the present invention, the sound generated during mechanical fastening, especially self-piercing riveting process, is measured, and unique identification data is captured from the measured acoustic data to evaluate the presence or absence of defects and quality of the mechanical fastening. .

In particular, the transformed data obtained by Fourier transforming the measured acoustic data is divided into a plurality of frequency sections, data preprocessing is performed including extracting feature data within the sections, and this is used as learning data to improve the evaluation precision of the fastening process positive and negative. Reliability and data processing speed can be increased.

Using the above learning model, non-destructive testing as well as real-time testing can be performed by measuring and analyzing the sound generated during the fastening process in real time.

Effects according to embodiments of the present invention are not limited to the contents exemplified above, and further various effects are included in the present specification.

1 is a hardware configuration diagram of a mechanical fastening facility according to an embodiment of the present invention.
Figure 2 is a schematic diagram of the mechanical fastening equipment of Figure 1.
Figure 3 is a schematic diagram showing the self-piercing riveting process in sequence.
Figure 4 is a flowchart showing an artificial intelligence-based learning method according to an embodiment of the present invention.
Figure 5 is a flowchart showing an artificial intelligence-based learning method according to another embodiment of the present invention.
Figure 6 is a flowchart showing a mechanical fastening process and a quality evaluation method of mechanical fastening according to an embodiment of the present invention.
Figures 7 and 8 are images showing Example 1.
Figure 9 is a flowchart showing a method for evaluating the quality of mechanical fastening according to another embodiment of the present invention.
Figures 10 and 11 are images showing Example 2.

The advantages and features of the present invention and methods for achieving them will become clear by referring to the embodiments described in detail below along with the accompanying drawings. However, the present invention is not limited to the embodiments disclosed below and will be implemented in various different forms, and only the embodiments serve to ensure that the disclosure of the present invention is complete, and those skilled in the art It is provided to fully inform the person of the scope of the invention, and the present invention is only defined by the scope of the claims. That is, various changes may be made to the embodiments presented by the present invention. The embodiments described below are not intended to limit the embodiments, but should be understood to include all changes, equivalents, and substitutes therefor.

Unless otherwise defined, all terms (including technical and scientific terms) used in this specification may be used with meanings that can be commonly understood by those skilled in the art to which the present invention pertains. Additionally, terms defined in commonly used dictionaries are not interpreted ideally or excessively unless clearly specifically defined.

As used herein, 'and/or' includes each and every combination of one or more of the mentioned items. Additionally, the singular form also includes the plural form unless specifically stated in the phrase. As used herein, 'comprises' and/or 'comprising' do not exclude the presence or addition of one or more other components in addition to the mentioned components. The numerical range expressed using 'to' indicates a numerical range that includes the values written before and after it as the lower limit and upper limit, respectively. ‘About’ or ‘approximately’ means a value or numerical range within 20% of the value or numerical range stated thereafter.

In this specification, when referring to components, ordinal modifiers such as 'first component', 'second component', '1-1 component', etc. are used to distinguish one component from another component. It just happens. Accordingly, the first component referred to below may be referred to as the second component within the scope of the technical idea of the present invention. For example, what is referred to as a first component in one embodiment may be referred to as a second component in another embodiment. For another example, what is referred to as a first element in the description of the invention may be referred to as a second element in the claims.

In this specification, the term 'computer' refers to all types of hardware devices that include at least one processor and perform specified instructions. Unless otherwise defined, a computer can be understood to also include software components implemented in hardware devices. Each step described below may be understood as being performed by a processor of a computer or computer device. Of course, in some embodiments, at least some of the steps may be performed by different computer devices or processors.

Hereinafter, the present invention will be described in detail with reference to the attached drawings.

1 is a hardware configuration diagram of a mechanical fastening facility according to an embodiment of the present invention. Figure 2 is a schematic diagram of the mechanical fastening equipment of Figure 1. Figure 3 is a schematic diagram showing the self-piercing riveting process in sequence.

Referring to FIGS. 1 to 3, the mechanical fastening device 11 (or mechanical fastening device, or riveting device, or artificial intelligence learning device, or mechanical fastening monitoring or quality evaluation device) according to the present embodiment is a processor (100). ), a memory 200, an acoustic measurement unit 600, a process unit 500, and a storage 300, and may further include an input/output unit 420 and a communication unit 410.

The processor 100 may implement the operations and functions of the mechanical fastening device 11 based on instructions according to software resident in the memory 200. For example, the processor 100 may perform an operation on at least one application or program to execute a method according to embodiments of the present invention. Examples of the processor 100 include a Central Processor Unit (CPU), Micro Processor Unit (MPU), Micro Controlling Unit (MCU), Graphic Processing Unit (GPU), Application-Specific Integrated Circuit (ASIC), other chipsets, Examples include logic circuits and/or data processing devices.

The memory 200 may be loaded with software. For example, memory 200 may load software (or computer program) from storage 300 . When software is loaded into the memory 200, the processor 100 may execute one or more instructions constituting the software to perform a method, operation, and/or function. The memory 200 may include read-only memory (ROM), random access memory (RAM), flash memory, a memory card, a storage medium, and/or other storage devices. Memory 200 may be internal or external to processor 100 and may be connected to processor 100 by various well-known means. For example, the processor 100 and memory 200 may together form a programmable logic controller (PLC), a PC (C3, C++), a self-program, and/or a combination thereof.

The storage 300 may store application programming interfaces, libraries, files, resource files, etc. necessary for executing software that implements a control method for operating the mechanical fastening equipment 11 according to the present invention. Meanwhile, the software residing in the memory 200 or stored in the storage 300 may be a computer program recorded on a recording medium to perform or execute a control method to be described later. For example, the storage 300 may be configured to include a known computer-readable recording medium, such as a read-only memory (ROM) or a hard disk.

Software resident in memory 200 or stored in storage 300 includes one or more instructions for evaluating or monitoring the quality of mechanical fastening, which will be described later, and/or performing or executing a mechanical fastening process, or is recorded on a computer recording medium. It can be a program. When the present invention is implemented in software, the method may be implemented as modules, such as functions, processes, etc. The module may reside in memory 200 and be executed by processor 100 .

The various components shown in FIG. 1 may be implemented by various means, such as hardware, firmware, software, or a combination thereof. In the case of implementation by hardware, an embodiment of the present invention includes one or more Application Specific Integrated Circuits (ASICs), Digital Signal Processors (DSPs), Digital Signal Processing Devices (DSPDs), Programmable Logic Devices (PLDs), and FPGAs ( Field Programmable Gate Arrays), processors, controllers, microcontrollers, microprocessors, etc. Or, in the case of implementation by firmware or software, an embodiment of the present invention is implemented in the form of a module, procedure, function, etc. that performs the functions or operations described above, and is recorded on a recording medium readable through various computer means. It can be. Here, the recording medium may include program instructions, data files, data structures, etc., singly or in combination. Program instructions recorded on the recording medium may be those specifically designed and constructed for the present invention, or may be known and available to those skilled in the art of computer software. For example, recording media include magnetic media such as hard disks, floppy disks and magnetic tapes, optical media such as CD-ROM (Compact Disk Read Only Memory) and DVD (Digital Video Disk), and floptical media. It includes magneto-optical media such as floptical disks, and hardware devices specially configured to store and execute program instructions such as ROM, RAM, flash memory, etc. Examples of program instructions may include machine language code such as that created by a compiler, as well as high-level language code that can be executed by a computer using an interpreter, etc. Such hardware devices may be configured to operate as one or more software to perform the operations of the present invention, and vice versa.

The software may include various information necessary to provide a learning method for learning an artificial intelligence model, a mechanical fastening monitoring method, a mechanical fastening method, etc., which will be described later, or one or more instructions for performing the same.

A data bus (BUS) may provide communication capabilities between components of a computing device. In other words, the data bus (BUS) is connected to the processor 100, memory 200, communication unit 410, input/output unit 420, storage 300, and other components to be described later, so that data is transmitted between each component. It can serve as a moving passage for delivering.

Communication unit 410 may include a communication interface. Although not shown in the drawing, when the mechanical fastening equipment 11 according to this embodiment requires network communication with a server (not shown), the communication unit 410 is a base band circuit for processing wired and wireless signals with them. may include.

When a user wants to directly control the mechanical fastening facility 11, the input/output unit 420 includes input devices such as buttons required for operation input, and output devices such as a display for displaying the current status of the mechanical fastening facility 11. It can be included.

Since the communication unit 410 and the input/output unit 420 may be known, detailed descriptions thereof will be omitted.

The process unit 500 may refer to a component for performing a mechanical fastening, for example, a riveting process. In an exemplary embodiment, when the mechanical fastener is a self-piercing rivet, process unit 500 may include a punch (P) and an anvil die (D). Generally, the self-piercing riveting process may be performed in the following order: clamping, piercing, flaring, and releasing.

Clamping involves interposing a joining object 100, for example, an upper piece 110 and a lower piece 120, between an anvil die (D) and a punch (P), aligning them, and then fixing them. It can mean. Figure 3 illustrates the case where the joining object 100 includes only two upper pieces 110 and the lower piece 120, but the present invention is not limited thereto and the joining object 100 may be composed of three or more pieces. It may be possible.

The upper piece 110 and the lower piece 120 may each include a metallic material or a non-metallic material. Examples of metal materials include iron (Fe), chromium (Cr), nickel (Ni), aluminum (Al), magnesium (Mg), or alloys thereof. Examples of non-metallic materials include carbon fiber reinforced plastic (CFRP).

Piercing may refer to the step of at least partially penetrating the rivet (R) into the joint object 100 using a punch (P). Pressing of the rivet (R) may be performed through a punch (P).

Flaring may refer to a step in which the rivet R penetrates the joint object 100 and changes its shape. For example, the shank of the rivet R may open as if exploding and deform the shape of a ductile piece, for example, the lower piece 120. The deformed shape may be modified to match the shape of the upper surface of the anvil die (D). Figure 3 illustrates a convexly protruding base surface of the anvil die (D), but the present invention is not limited thereto. The upper piece 110, the lower piece 120, whose shape is changed during the flaring process, and the rivet R are engaged with each other, and the upper piece 110 and the lower piece 120 can be firmly fixed.

Lastly, releasing may refer to the step of releasing the fixed state by moving the punch (P) upward after the mechanical fastening is completed.

The acoustic measurement unit 600 may be located adjacent to the punch (P) and the anvil die (D). That is, the sound measurement unit 600 is located adjacent to the process space between the punch (P) and the anvil die (D), and can measure sound generated in the process space. The acoustic measurement unit 600 may include an acoustic sensor such as a microphone.

The acoustic signal measured by the acoustic measurement unit 600 may be expressed as the size of the acoustic signal over time. The size of the sound signal may be expressed in terms of pressure or decibel (dB). For example, measured raw data can be collected and stored in a format where the horizontal axis is time and the vertical axis is the size of the sound signal.

In addition, the sound signal collected by the sound measurement unit 600 may be converted into the size of the sound signal in terms of frequency (Hz) by a sound processing unit of a computer device, such as the processor 100. For example, the acoustic signal measured by the acoustic measurement unit 600 may be Fourier transformed. A detailed example of the Fourier transform is the Fast Fourier Transform (FFT). At this time, the size of the converted sound signal can be expressed as pressure, decibel, or amplitude. That is, the processor 100 can convert the size of the signal over process time into the size of the signal for each frequency band.

In other words, the method for monitoring or evaluating the mechanical fastening quality according to this embodiment is performed based on the acoustic signal measured by the acoustic measurement unit 600, or Fourier transforms the acoustic signal collected by the acoustic measurement unit 600. It can be performed based on converted data.

The mechanical fastening equipment 11 according to the present invention measures the sound generated during the mechanical fastening process using the process unit 500, that is, the self-pierce riveting process, and determines the quality of the performed process using a previously learned learning model. Or, you can monitor it. Meanwhile, the processing unit 500 and the acoustic measurement unit 600 may be timing controlled by the processor 100.

Hereinafter, the artificial intelligence learning method according to the present invention will be described in more detail. Figure 4 is a flowchart showing an artificial intelligence-based learning method according to an embodiment of the present invention.

Referring further to FIG. 4, the artificial intelligence learning method or the method of building a learning model according to this embodiment includes measuring an acoustic signal using an acoustic measurement unit 600 (S110), and separating the collected acoustic signals according to time. Step (S210), obtaining and storing converted data by Fourier transforming at least some of the collected acoustic signals using a Fourier transform unit, obtaining and storing converted data by imaging the converted data. It may include a step S410 and a step S510 of learning using image data as training data.

The sound signal measured in the step of measuring the sound signal (S110) may be expressed as the size of the sound signal over time, as described above. For example, the measured sound signal may be in a form where the horizontal axis is time and the vertical axis is the size of the sound signal, for example, pressure (Pa) or decibel (dB).

And the measured raw data, that is, the acoustic signal, can be separated according to time (S210). In other words, the sound signal of a specific time period can be extracted. Separation or extraction of the acoustic signal may be performed by logical components related to the acoustic signal time extraction unit of software, but the present invention is not limited thereto.

Generally, in the self-piercing riveting process, clamping, piercing, flaring, and releasing processes may be performed sequentially, and an acoustic signal of a predetermined size may be detected for each step. The method of performing the step of separating the acoustic signal is not particularly limited, but, for example, it may be performed based on the time when the acoustic measurement unit 600 starts measuring acoustic sound greater than a predetermined standard size and the time when acoustic measurement ends. .

As a non-limiting example, an acoustic signal of a certain size measured for about 0.2 milliseconds (ms) in the period from 1 to 2 seconds after the start of the acoustic measurement may be generated during clamping. Additionally, an acoustic signal of a certain size measured for about 0.5 ms in the period from 2 to 2.5 seconds after the start of the acoustic measurement may have occurred during piercing. An acoustic signal of a certain size measured for about 0.5 ms in the period from 2.5 to 3 seconds after the start of the acoustic measurement may have occurred during flaring. In addition, an acoustic signal of a predetermined size measured for about 0.2 ms immediately after flaring, for example, in the period from 3 to 3.5 seconds after the start of acoustic measurement, may have occurred during releasing. In this way, the sound can be separated based on the time section, or the sound can be sequentially separated from sound signals detected with a size above or below a certain size, or within a predetermined range.

Hereinafter, the case where an acoustic signal generated during a piercing process in a period of 2 seconds to 2.5 seconds is extracted (S210) and Fourier transform, which will be described later, is performed will be described as an example, but the present invention is not limited thereto. Unlike this embodiment, conversion data may be collected based on acoustic signals generated during the entire process section, but in terms of efficiency of data processing and conversion, it may be preferable to use only acoustic signals that can be used as feature data as in this embodiment. there is. The inventors of the present invention confirmed that when learning is performed based on the acoustic signal generated during the piercing section compared to the acoustic signal generated in other process sections, higher precision is achieved compared to the acoustic signal generated in other process sections, and thus completed the present invention. It has arrived.

Next, the extracted acoustic signal can be converted to collect conversion data (S310). Conversion of the acoustic signal may be performed by logical components related to the Fourier transform unit of software, but the present invention is not limited thereto.

A detailed example of the Fourier transform is the Fast Fourier Transform (FFT). Fourier transformed data can be expressed as the size of an acoustic signal according to frequency. For example, the converted sound signal may be in a form where the horizontal axis is frequency and the vertical axis is the size of the sound signal, such as pressure, decibel, or dimensionless amplitude.

Next, the converted data can be converted and saved in image form (S410). That is, image data can be extracted from acoustic signal data. Imaging of converted data may be performed by logical components related to the image conversion unit of software, but the present invention is not limited thereto.

As described above, data collected by the first acoustic measurement unit 600, that is, data on the size of the acoustic signal over time, may not provide sufficient feature points to perform artificial intelligence learning. However, when Fourier transforming an acoustic signal in a specific time period as in this embodiment, as a non-limiting example, peaks in about 10 or more, about 15 or more, or about 20 or more specific frequency bands are used as feature data. This can increase the precision of learning and quality evaluation.

And learning can be performed using the acquired image data (S510). At this time, various artificial intelligence learning models, such as machine learning models or deep learning models, can be used as the learning model using image data. For example, in finding various relationships and patterns from collected data, individual models such as R-convolutional neural network (region-based convolutional neural network) or Mask-R-CNN are designed with depth and structure suitable for class classification. A region partition model may be utilized.

The method of providing classes is not particularly limited, but as a non-limiting example, 13 classes including various normal and bad conditions as shown in Table 1 below can be used.

class cross section image explanation One normal 2 Rivet leg length. -0.5mm 3 Rivet leg length. +0.5mm 4 Missalignment 1mm 5 Missalignment 2mm 6 HD2 Rivet+0.5mm 7 K.A.Die 8 DZ Die 9 Gap 1mm 10 Gap 2mm 11 Tilt 5 12 Tilt 10 13 Tilt 15

Hereinafter, other embodiments of the present invention will be described. However, descriptions of configurations that are substantially the same or extremely similar to the above-described embodiments will be omitted, and this will be easily understood by those skilled in the art from the attached drawings.

Figure 5 is a flowchart showing an artificial intelligence-based learning method according to another embodiment of the present invention.

Referring to FIG. 5, the artificial intelligence learning method or the method of building a learning model according to this embodiment includes the steps of measuring an acoustic signal using the acoustic measurement unit 600 (S110), and separating the collected acoustic signals according to time. A step (S210), and a step (S310) of acquiring and storing converted data by Fourier transforming at least some of the collected acoustic signals using a Fourier transform unit, and storing some of the frequency-specific sections of the transformed data. It may include a step of separating (S420) and a step of learning using the signal size for each section as learning data (S520).

Since the step of measuring the sound signal (S110), the step of extracting the sound signal (S210), and the step of Fourier transforming (S310) have been described previously, overlapping descriptions will be omitted.

Next, the Fourier transformed data is divided into a plurality of sections for each frequency section (S420). Unlike this embodiment, in terms of data processing and efficiency, it may be based on the signal of the entire frequency section. However, as in this embodiment, it is better to extract and divide frequency sections that can be used as feature data and perform learning on each of them. It was confirmed that higher precision was achieved compared to other cases, and the present invention was completed.

In an exemplary embodiment, the step of extracting and dividing the frequency section (S420) may include extracting one or more of the following five sections. Division of the frequency section may be performed by logical components related to the frequency section division unit of software, but the present invention is not limited thereto.

1) 100Hz to 150Hz, 2) 150Hz to 400Hz, 3) 400Hz to 550Hz, 4) 550Hz to 2,000Hz, 5) 2,000Hz to 2,550Hz

And learning can be performed using the acquired signal size for each frequency section (S520). At this time, one or more combinations of previously extracted and stored frequency sections can be used.

For example, learning can be performed using information included in any one of the five sections. For another example, learning may be performed using information included in two or more of the five sections.

At this time, when using multiple frequency section data for learning, data for each may be used, or correlation between data within a plurality of frequency sections may be used.

For example, a first section (e.g., 100 Hz to 150 Hz) and a different second section (e.g., 400 Hz to 550 Hz) are used to learn, and if they are all within the learned numerical range, they can be classified as belonging to a specific class. there is.

For another example, by comparing the characteristics of data (e.g., peak area, area of range section, etc.) in a first interval (e.g., 100 Hz to 150 Hz) with the characteristics of data in another second interval (e.g., 400 Hz to 550 Hz), It can be learned based on correlations such as size or ratio, and if the correlation is within the learned numerical range, it can be classified as belonging to a specific class.

As described above, when each section is used as learning data or the correlation between feature data of a plurality of sections is used as learning data, a plurality of consecutive sections among the five sections described above may be set as one section. For example, any three sections (e.g., 100Hz to 150Hz, 150Hz to 400Hz, 400Hz to 550Hz) are set as one first section, and one or more other sections (e.g., 2,000Hz to 2,550Hz) are set as the second section. It can also be set to .

In this step (S520), the Fourier transform data values extracted to be used as learning data or for correlation with other sections include the number of peaks in the section, the value of each peak, the area of each peak, and the It may be the average value of the peak, the standard deviation of the peak value within each section, the area of each section, and/or the tendency to increase or decrease the signal size at the peak. Here, the peak forms an extreme value before and after it, and may mean that the signal size at a predetermined frequency before and after and the rate of change of the signal size at that frequency are greater than a predetermined standard.

In more detail, the number of peaks within a section may mean the number of peaks counted within the corresponding frequency section.

Additionally, the value of each peak may mean the signal size corresponding to the peak. In some embodiments, the values of peaks may be simplified through certain preprocessing procedures. As a non-limiting example, if the signal size of the peak in a certain frequency range is sequentially 2.143, 2.174, 2.781, 3,415, and 1,742, these can be expressed to the first decimal place by rounding or truncating, or expressed as an integer. .

The average value and standard deviation of the peaks within each section may mean the average and standard deviation of the signal magnitude values of the peaks in the corresponding frequency section.

Additionally, the area of each peak may mean the area of the peak within a range before and after a predetermined frequency from the frequency value representing the extreme value. For example, if a peak is expressed at 1,540 Hz, the frequency area in a predetermined frequency range of ±5 Hz can be calculated. As a non-limiting example, the integrated area in the section from 1,535 Hz to 1,545 Hz may correspond to the area of the 1,540 Hz peak.

Additionally, the area of each section may mean the total integrated area in a specific frequency section, for example, 400Hz to 550Hz.

In other words, the Fourier transformed data can be preprocessed to extract quantitative data related to peaks and then used to perform learning. The learning model may use a known artificial intelligence learning model, for example, ANN, DNN, etc. The provided classes are explained in Table 1, so redundant explanations are omitted.

In an exemplary embodiment, learning may utilize the area of one or more peaks occurring in the range 100 Hz to 550 Hz, or the area of a range interval, and the area of one or more peaks occurring in the range 2,000 Hz to 2,550 Hz, or the area of a range interval. You can.

In supplementary, additional and/or alternative embodiments, learning includes the value of one or more peaks, or number of peaks, occurring in the range from 100 Hz to 550 Hz, and the value of one or more peaks, or number of peaks, occurring in the range from 2,000 Hz to 2,550 Hz. number can be used.

In supplementary, additional and/or alternative embodiments, learning is performed on the area of one or more peaks occurring in the range from 100 Hz to 150 Hz, or an area of a section of said range, and the area of one or more peaks occurring in the range of 400 Hz to 550 Hz, or said range. The area of the section can be used.

In a supplementary, additional and/or alternative embodiment, learning may include the value or number of peaks of one or more peaks occurring in the range of 100 Hz to 150 Hz, and the value or number of peaks of one or more peaks occurring in the range of 400 Hz to 550 Hz. Available.

In supplementary, additional and/or alternative embodiments, learning is performed by measuring the area of one or more peaks occurring in the range from 400 Hz to 550 Hz, or the area of a section of the range, and the area of one or more peaks occurring in the range from 550 Hz to 2,000 Hz, or The area of the range section can be used.

Figure 6 is a flowchart showing a mechanical fastening process and a quality evaluation method of mechanical fastening according to an embodiment of the present invention.

Referring further to FIG. 6, the mechanical fastening process and the quality evaluation method of the mechanical fastening process according to the present embodiment include measuring an acoustic signal using an acoustic measurement unit 600 (S610), and measuring the collected acoustic signal over time. A step of separating (S620) and Fourier transforming at least some of the collected acoustic signals using a Fourier transform unit to obtain and store transformed data (S630), and evaluating the stored Fourier transform data and the learned evaluation. A step (S691) of evaluating quality by comparing models may be further included.

The step of measuring the acoustic signal (S610), the step of extracting the acoustic signal (S620), and the step of Fourier transforming (S630) can be performed in substantially the same way using the same components as the artificial intelligence learning method described above. .

Then, the converted data can be compared with the previously learned evaluation model to classify the class and determine whether it is normal (S691).

The present invention will be described in more detail below through specific examples.

<Example 1>

Learning was performed to evaluate the quality of self-piercing rivets using the learning model shown in Figure 7. The learning method was as shown in Figure 5 and Table 1. For learning, the section was divided into 5 sections and all of them were used. In Table 1, classes 2 to 13 can be understood as defective. Then, using the learned model, the self-piercing riveting process corresponding to classes 1 to 13 was performed 20 times, and the results and error matrix regarding classification accuracy were created and shown in FIG. 8.

As an acoustic measurement unit, a GRAS 46AE microphone from Bruel&Kjaer was prepared. The NI-9234 module and Matlab were used as the sound processing unit. In addition, as a mechanical fastener, FM type 090-2-018 product was used as the anvil die, C-type (Φ5.3×6.5) rivet was used, and the pressure of the punch unit was set to 45kN.

Referring to Figure 8, you can see the results of classifying 13 classes with 100% precision in 20 tests.

Meanwhile, the present invention provides, in addition to a method for evaluating the quality of a mechanical fastening process through artificial intelligence learning, a method for evaluating the quality of a mechanical fastening process performed by a computer device. Below, this will be explained.

Figure 9 is a flowchart showing a method for evaluating the quality of mechanical fastening according to another embodiment of the present invention.

Referring to FIG. 9, the mechanical fastening process and the quality evaluation method of the mechanical fastening process according to the present embodiment include measuring an acoustic signal using an acoustic measurement unit 600 (S610), separating the collected acoustic signals according to time. A step (S620) of obtaining and storing transformed data by Fourier transforming at least some of the collected acoustic signals using a Fourier transform unit, and obtaining and storing transformed data, and the signal size for each section of the stored Fourier transformed data. A step (S692) of evaluating quality through comparison may be further included.

Since the step of measuring the sound signal (S610), the step of extracting the sound signal (S620), and the step of Fourier transforming (S630) have been described previously, redundant descriptions will be omitted.

The step (S692) of evaluating quality through comparison of signal size, etc. may be performed using logical components related to a comparison judgment unit of a computer device, for example, software, but the present invention is not limited thereto. The inventors of the present invention discovered that there was a difference in signal characteristics for each frequency band when the result of mechanical fastening, such as self-piercing riveting, had excellent fastening strength and when it did not, and took note of this to complete the present invention.

Although not shown in the drawing, the step of evaluating quality (S692) may include extracting and dividing a frequency section and comparing and judging them. Division of the frequency section may be performed by logical components related to the frequency section division unit of software, but the present invention is not limited thereto. Extracting and dividing the frequency section may include extracting one or more of the following five sections.

1) 100Hz to 150Hz, 2) 150Hz to 400Hz, 3) 400Hz to 550Hz, 4) 550Hz to 2,000Hz, 5) 2,000Hz to 2,550Hz

In an exemplary embodiment, the Fourier transform of the acoustic signal during piercing results in the ratio of the sum of the amplitudes of the signals occurring in the range of 100 Hz to 550 Hz (G1) to the sum of the amplitudes of the signals occurring in the range of 2,000 Hz to 2,550 Hz (G2). If (G1/G2) is 10 or more, preferably 12 or more, it is judged to be normal. Otherwise, if it is not, preferably less than 5.0, or about 4.5 or less, it can be judged as defective.

In this embodiment, the sum of signals occurring in a specific frequency section can be understood as an integral value for the corresponding frequency section in a graph with the horizontal axis as the frequency range and the vertical axis as the signal size.

In a complementary, additional and/or alternative embodiment, the Fourier transform of the acoustic signal during piercing results in a magnitude (H1) of the signal having a maximum peak in the range of 100 Hz to 550 Hz, with a maximum peak in the range of 2,000 Hz to 2,550 Hz. If it is more than 3 times the size of the signal (H2), preferably more than 4 times, it can be judged as normal. Otherwise, if it is less than 2 times, or preferably less than about 1.5 times, it can be judged as defective.

In this embodiment, the size of a signal generated at a specific frequency value can be expressed as the size of the signal at that frequency value, that is, a value on the vertical axis, in a graph with the horizontal axis as the frequency band and the vertical axis as the signal size.

The inventors of the present invention focused on the fact that signals in the high frequency band during piercing, for example, a frequency band of 2,000 Hz or more, and signals in the low frequency band, for example, the frequency band of 550 Hz or less, show a specific difference in behavior depending on the normal presence or absence of self-piercing rivet fastening. The invention has been completed.

In other words, the present invention is not limited to any theory, but when the rivet fastening is normally performed, for example, no acoustic signal is substantially generated at 2,000 Hz to 2,550 Hz, or about 2,200 Hz to 2,500 Hz, or about 2,300 Hz to 2,450 Hz. On the other hand, if the rivet fastening is performed abnormally, an acoustic signal of a predetermined size may be generated in the frequency section.

Also, in a complementary, additional and/or alternative embodiment, the Fourier transform of the acoustic signal during piercing results in the magnitude of the signal occurring in the range of 400 Hz to 550 Hz relative to the sum (G2) of the magnitude of the signal occurring in the range of 100 Hz to 150 Hz. If the ratio (G1/G2) of the sum (G1) of there is.

In a complementary, additional and/or alternative embodiment, the Fourier transform of the acoustic signal during piercing results in a magnitude (H1) of the signal with a maximum peak in the range of 100 Hz to 150 Hz, which is greater than or equal to that of the signal with a maximum peak in the range of 400 Hz to 550 Hz. If it is 2.5 times the size (H2), preferably 2.8 times or more, it can be judged as normal. Otherwise, if it is preferably 1.8 times or less, or about 1.5 times or less, it can be judged as defective.

The inventors of the present invention focused on the fact that, among signals in the low frequency band during piercing, for example, frequencies in the frequency band below 550 Hz, ultra-low frequency signals and low-mid frequency signals show specific behavioral differences depending on whether self-piercing rivet fastening is normal or not, and developed the present invention. It has reached completion.

In other words, the present invention is not limited to any theory, but when the rivet fastening is performed normally, an acoustic signal is mainly generated at 100 Hz to 150 Hz, and a weak acoustic signal is generated at 400 Hz to 550 Hz, whereas when the rivet fastening is performed abnormally. An acoustic signal larger than a certain size may be generated at 400Hz to 550Hz, or a signal larger than that at 100Hz to 150Hz may be generated.

Furthermore, in a complementary, additional and/or alternative embodiment, the result of the Fourier transform of the acoustic signal during piercing is the sum of the amplitudes of the signals occurring in the range of 550 Hz to 2,000 Hz relative to the sum of the amplitudes of the signals occurring in the range of 400 Hz to 550 Hz. If the ratio is 0.5 or less, preferably 0.3 or less, it can be judged as normal. Otherwise, if it is not, preferably 0.8 or more, or about 1.0 or more, it can be judged as defective.

The inventors of the present invention have discovered a specific difference in behavior depending on the normal presence or absence of self-piercing rivet fastening in signals in the low-frequency band during piercing, for example, a frequency band of 550 Hz or less and in a mid-frequency band, for example, in a frequency band ranging from 550 Hz to 2,000 Hz. Focusing on what is shown, the present invention was completed.

In other words, the present invention is not limited to any theory, but when the rivet fastening is performed normally, the sound signal does not substantially occur at 550 Hz to 2,000 Hz, or if it does occur, it is very weak, whereas when the rivet fastening is abnormal, the An acoustic signal of a certain size may be generated in a frequency section.

The present invention will be described in more detail below through specific examples.

<Example 2>

Monitoring of mechanical fastening quality was performed in the same manner as in Figure 9. The mechanical fastening method was the same as in Example 1 described above. The sound signal over time measured when the rivet is fastened normally, the Fourier transform results of the signal generated during the piercing process, and the cross-sectional image of the rivet are shown in Figure 10. In addition, in the case where the rivet is fastened abnormally, the sound signal over time measured, the Fourier transform result of the signal generated during the piercing process, and the cross-sectional image of the rivet are shown in FIG. 11.

Referring to Figures 10 and 11, it can be seen that in the case of a defect, an acoustic signal with a relatively very large peak is detected in a graph of the size of the sound signal over time compared to the normal case. In particular, there is a clear difference in the detection signal during piercing between normal and defective cases, and it can be seen that in the defective case, an acoustic signal of more than 250Mpa was detected, whereas in the normal case, this was not the case.

In addition, in the Fourier transform data graph, it can be seen that the sum of the acoustic signals in the defective case is relatively larger over the entire section compared to the normal case.

Specifically, in the normal case, the sum of the signal sizes generated in the range of 2,000 Hz to 2,550 Hz is significantly smaller than in the defective case. In addition, the sum of the magnitudes of signals generated in the range of 100Hz to 150Hz in normal and defective cases was similar. Additionally, in the normal case, the sum of the signal sizes generated at 400Hz to 550Hz is significantly smaller than in the defective case. That is, the sum of signal sizes from 100 Hz to 550 Hz, or 100 Hz to 150 Hz, can be used as a standard, and a predetermined relationship with the sum of signal sizes in other frequency bands can be established based on this.

In the normal case, the ratio of the sum of the signals in the 100 Hz to 550 Hz range to the sum of the signals in the 2,000 Hz to 2,550 Hz range was about 20, but in the defective case it was about 2.5.

In addition, in normal cases, the size of the signal with the maximum peak in the range of 100 Hz to 550 Hz was about 20 times that of the signal with the maximum peak in the range of 2,000 Hz to 2,550 Hz, but in the case of defects, it was about twice that.

Additionally, in the normal case, the ratio of the sum of the signals in the 400Hz to 550Hz range to the sum of the signals in the 100Hz to 150Hz range was about 1.7, but in the defective case it was 0.9.

And in the normal case, the size of the signal with the maximum peak in the 100Hz to 150Hz range was about 2.8 times that of the signal with the maximum peak in the 400Hz to 550Hz range, but in the defective case, it was about 1.6 times the size.

In addition, in the normal case, the ratio of the sum of the amplitudes of the signals occurring in the range of 550 Hz to 2,000 Hz to the sum of the amplitudes of the signals occurring in the range of 400 Hz to 550 Hz was about 0.5, but in the defective case it was 0.9.

Although the description has been made above with a focus on embodiments of the present invention, this is merely an example and does not limit the present invention, and those skilled in the art will be able to understand the present invention without departing from the essential characteristics of the embodiments of the present invention. It will be apparent that various modifications and applications not exemplified above are possible.

Therefore, the scope of the present invention should be understood to include changes, equivalents, or substitutes of the technical ideas exemplified above. For example, each component specifically shown in the embodiments of the present invention can be modified and implemented. And these variations and differences in application should be construed as being included in the scope of the present invention as defined in the appended claims.

Additionally, each component on the configuration diagram or block diagram of the attached drawings may be implemented with hardware such as software, field-programmable gate array (FPGA), or application specific integrated circuit (ASIC). However, each component on the configuration diagram or block diagram may be implemented in an addressable storage medium as well as software and hardware, and may be configured to execute one or more processors.

Each component on a configuration diagram or block diagram may mean a module, segment, or part of code that includes one or more executable instructions for executing a specified logical function. Therefore, the functions provided by the components on the configuration diagram or block diagram may be implemented by a plurality of more detailed components, or the multiple components on the configuration diagram or block diagram may be implemented by one integrated component. Of course.

That is, within the scope of the purpose of the present invention, each component can be operated by selectively combining one or more of them. In addition, all components may be implemented as independent hardware, and some or all of the components may be selectively combined to have a program module that performs some or all of the combined functions in one or more pieces of hardware. It may also be implemented as a computer program. The codes and code segments that make up the computer program can be easily deduced by those skilled in the art. Such a computer program can be stored in a computer-readable recording medium or storage medium and can be read and executed by a computer to implement embodiments of the present invention. Examples of the recording medium include magnetic recording media, optical recording media, and the like.

11: Mechanical fastening equipment
100: processor
200: memory
300: Storage
410: communication unit
420: input/output unit
500: Process unit
600: Acoustic measurement unit

Claims (10)

A method performed by a computing device, comprising:
A mechanical device that measures acoustic signals generated during the self-piercing riveting process performed in the order of clamping, piercing, and flaring, and uses data based on the conversion data obtained by Fourier transforming at least some of the measured acoustic signals as learning data. Artificial intelligence learning method for predicting joint quality. According to paragraph 1,
The learning data is an artificial intelligence learning method that uses image conversion data with frequency on the horizontal axis and amplitude on the vertical axis. According to paragraph 1,
The learning data divides the converted data into a plurality of frequency sections, and includes the number of peaks in each section, the value of each peak, the area of each peak, the average value of the peak values within each section, and the standard deviation of the peak values within each section. and an artificial intelligence learning method that uses one or more of the areas of each section. According to paragraph 3,
The learning data is at least
The area of one or more peaks occurring in the range from 100 Hz to 550 Hz, or the area of a section of the range, and
An artificial intelligence learning method that uses the area of one or more peaks occurring in the range of 2,000 Hz to 2,550 Hz, or the area of the range section. According to paragraph 3,
The learning data is at least
The value of one or more peaks, or the number of peaks, occurring in the range from 100 Hz to 550 Hz, and
An artificial intelligence learning method that uses the value of one or more peaks or the number of peaks occurring in the range of 2,000 Hz to 2,550 Hz. According to paragraph 3,
The learning data is at least
The area of one or more peaks occurring in the range from 100 Hz to 150 Hz, or the area of a section of the range, and
An artificial intelligence learning method that uses the area of one or more peaks occurring in the range of 400Hz to 550Hz, or the area of the range section. According to paragraph 3,
The learning data is at least
The value of one or more peaks, or the number of peaks, occurring in the range from 100 Hz to 150 Hz, and
An artificial intelligence learning method that uses the value of one or more peaks or the number of peaks occurring in the range of 400Hz to 550Hz. According to paragraph 3,
The learning data is at least
The area of one or more peaks occurring in the range from 400 Hz to 550 Hz, or the area of a section of the range, and
An artificial intelligence learning method that uses the area of one or more peaks occurring in the range of 550Hz to 2,000Hz, or the area of the range section. A self-piercing riveting device that performs clamping, piercing and flaring in that order,
An acoustic measurement unit that measures the size of an acoustic signal over time during the process;
a Fourier transform unit that performs Fourier transform on the sound signal; and
A riveting device comprising a quality evaluation unit that evaluates process quality by using the Fourier transformed data as an input to a learned learning model. A quality evaluation method of a self-piercing riveting process performed in the order of clamping, piercing and flaring, performed by a computing device,
A method of evaluating the quality of the riveting process by measuring acoustic signals generated during the riveting process and using the converted data obtained by Fourier transforming at least some of the measured acoustic signals as input data of the learned evaluation model.

Images