Inspection method and inspection device for workpiece in laser processing

Abstract

The inspection method includes: a step of acquiring a signal indicating a change in a component during a period corresponding to a processing time of each workpiece, the signal being generated by detecting, by an optical sensor, at least one of heat radiation, visible light, and reflected light generated by irradiating the workpiece with laser light; a step for calculating a feature amount indicating a feature of the signal during a predetermined period of the periods; a step for inputting the calculated feature quantity into a determination model for determining the surface roughness indicating the surface properties of the surface of the workpiece irradiated with the laser light, and determining the surface roughness of the workpiece; and a step for outputting the calculated predicted value of the surface roughness as an inspection result.

Description

Technical Field

[0001] The present disclosure relates to a method and apparatus for inspecting a workpiece in laser processing. Background Art

[0002] Patent Document 1 discloses a method for laser welding, which involves irradiating a workpiece with a pulsed laser beam to perform welding. This method determines the weld condition, such as whether the weld is good or bad, in the workpiece. In the method of Patent Document 1, the intensities of plasma light and reflected light emitted from the workpiece during laser welding are detected. Based on the detected light intensity within a pre-set extraction interval within a cycle corresponding to a single laser pulse, a pulse characteristic value is extracted for each laser pulse. The average value of the detected light intensity, the amount of change in differential processing, and the amplitude of the differential processing are calculated as each pulse characteristic value. The method of Patent Document 1 compares the lower or upper limit of each pulse characteristic value with a specified threshold value to determine the presence of welding defects based on the weld condition of each workpiece.

[0003] Prior art literature Patent Literature Patent Document 1: Japanese Patent Application Laid-Open No. 2000-153379 Summary of the Invention

[0004] In laser processing, such as welding, the condition of the workpiece can affect processing accuracy and finished quality. For example, in laser welding, variations in surface roughness can sometimes affect weld quality. Identifying the cause of such an effect requires detailed analysis, including factors other than surface roughness, such as cross-sectional observation of the processed portion. Furthermore, in situations where repeated processing is performed using equipment with high production volumes, measuring surface roughness during each process increases production time. Furthermore, surface roughness measurement requires a highly accurate measuring instrument, making it impractical to inspect the workpiece by measuring surface roughness before each process in production facilities.

[0005] The present disclosure provides an inspection device and an inspection method that can easily inspect the surface roughness of a workpiece during laser processing.

[0006] According to one embodiment of the present disclosure, a method for inspecting a workpiece during laser processing is provided.

[0007] The inspection method includes the following steps: a step of acquiring a signal generated by detecting at least one of the components of thermal radiation, visible light, and reflected light generated by irradiating a workpiece with a laser beam using a light sensor, the signal representing a change in the component in a time interval corresponding to a processing time for each workpiece; a step of calculating a characteristic quantity representing a characteristic of a signal in a predetermined interval within a time interval; a step of inputting the calculated feature value into a surface roughness determination model for determining surface properties of a surface of a workpiece irradiated with laser light, thereby determining the surface roughness of the workpiece; and This is a process of outputting the calculated predicted value of surface roughness as an inspection result.

[0008] The determination model is constructed based on training data including feature quantities calculated from signals of components detected by laser processing under each of a plurality of conditions that vary the surface roughness, in association with the surface roughness under each condition.

[0009] According to one embodiment of the present disclosure, an inspection device for a workpiece during laser processing is provided. The inspection device includes: an arithmetic circuit; and a communication circuit that receives a signal generated by a light sensor detecting at least one of the components of thermal radiation, visible light, and reflected light generated by irradiating the workpiece with laser light. The signal represents a change in the component within a time interval corresponding to the processing time of each workpiece. The arithmetic circuit obtains the signal via the communication circuit, calculates a feature quantity representing a characteristic of the signal within a predetermined interval within the time interval, inputs the calculated feature quantity into a determination model for determining the surface roughness representing the surface properties of the workpiece irradiated with laser light, determines the surface roughness of the workpiece, and outputs a predicted value of the calculated surface roughness as an inspection result. The determination model is constructed based on training data, which includes feature quantities calculated from signals of components detected by laser processing under each of multiple conditions that cause surface roughness to vary, in association with the surface roughness of each condition.

[0010] According to the inspection method and the inspection apparatus of the present disclosure, the surface roughness of a workpiece during laser processing can be easily inspected. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] Figure 1 This is a diagram showing an overview of the inspection system according to the first embodiment of the present disclosure.

[0012] Figure 2 This is a diagram illustrating the configuration of a laser processing device in an inspection system.

[0013] Figure 3 This is a diagram illustrating the configuration of a spectrometer in an inspection system.

[0014] Figure 4 This is a block diagram illustrating the configuration of an inspection device in an inspection system.

[0015] Figure 5This is a flowchart illustrating the inspection process in the inspection device.

[0016] Figure 6 This is a diagram for explaining signals obtained in the inspection device.

[0017] Figure 7 This is a diagram for explaining the relationship between the characteristic amount calculated by the inspection device and the surface roughness.

[0018] Figure 8 This is a flowchart illustrating a training process of a determination model used in an inspection process.

[0019] Figure 9 This is a diagram for explaining the training data of the judgment model.

[0020] Figure 10 This is a flowchart illustrating the process of generating training data. DETAILED DESCRIPTION

[0021] The following description of the embodiments will be described in detail with reference to the accompanying drawings as appropriate. However, unnecessary detailed descriptions may be omitted. For example, detailed descriptions of well-known matters and repeated descriptions of substantially identical structures may be omitted. This is to avoid unnecessary redundancy in the following description and to facilitate understanding by those skilled in the art. Furthermore, the inventors provide the accompanying drawings and the following description to facilitate a thorough understanding of the present disclosure by those skilled in the art, and the subject matter described in the claims is not limited to these.

[0022] (Implementation 1) In the first embodiment, as an example of using the inspection method and inspection apparatus of the present disclosure, an inspection system is described that detects components of light generated in laser processing for overlap welding and obtains signals based on the detected components to inspect the surface roughness of a workpiece.

[0023] 1. Structure use Figure 1 The inspection system according to the first embodiment will be described. Figure 1 It is a diagram showing an overview of the inspection system 100 according to this embodiment.

[0024] 1-1. System Overview Inspection system 100 includes a laser processing device 30 for performing laser processing for lap welding, a spectrometer 40 for detecting light components, and an inspection device 50. Laser processing workpiece 70, for example, is composed of metal. When irradiated with laser light 6, it generates thermal radiation in the near-infrared region due to a temperature rise, as well as luminescence or plasma luminescence (hereinafter referred to as "visible light"), which is primarily a visible light component and is inherent to the metal. Furthermore, a portion of laser light 6 that does not contribute to the processing is reflected as return light. Thus, when laser light 6 is irradiated from laser processing device 30 onto workpiece 70, thermal radiation, visible light, and reflected light are generated, for example, at molten zone 27 formed in workpiece 70 by melting of the metal.

[0025] The generated light is focused by the laser processing device 30, passes through the optical fiber 13 connecting the laser processing device 30 and the spectrometer 40, and is transmitted to the spectrometer 40. The light transmitted to the spectrometer 40 is split into components: thermal radiation, visible light, and reflected light. The components are detected by the optical sensor 22 of the spectrometer 40 and converted into signals. Upon receiving the signal from the spectrometer 40, the inspection device 50 determines the surface roughness of the workpiece 70 based on the received signal and outputs the determined surface roughness as the inspection result of the workpiece 70.

[0026] 1-2. Structure of Laser Processing Device Figure 2 1 is a diagram illustrating the configuration of a laser processing device 30 according to this embodiment. The laser processing device 30 includes a laser oscillator 1 , a laser transmission optical fiber 2 , a lens barrel 3 , a collimating lens 4 , condensing lenses 5 and 11 , a first reflecting mirror 7 , and a second reflecting mirror 8 .

[0027] Laser oscillator 1 supplies light for generating pulsed laser light 6 having a wavelength of approximately 1070 nanometers (nm). The light supplied from laser oscillator 1 is amplified while being transmitted through laser transmission optical fiber 2, passes through collimating lens 4 for obtaining a parallel beam, and is transformed into laser light 6, which then travels straight within lens barrel 3. Lens barrel 3 constitutes the processing head of laser processing device 30.

[0028] The laser beam 6 is reflected except for a portion that is transmitted by the first reflector 7, and is focused by the condenser lens 5. The laser beam is then irradiated onto, for example, a workpiece 70 secured to the scanning stage by the pressing jig 26. Thus, laser processing for lap welding of the workpiece 70 is performed. The wavelength of the laser beam 6 is not particularly limited to 1070 nm, but preferably a wavelength having a high absorption rate of the material is used.

[0029] When laser light 6 is irradiated, heat radiation from the workpiece 70, visible light from plasma emission, and reflected light from the laser light 6 are generated in the molten portion 27. These light components pass through the first reflector 7, are reflected by the second reflector 8, are focused by the condenser lens 11, and are then transmitted to the spectrometer 40 via the optical fiber 13. The laser processing device 30 of this embodiment further includes a light sensor 25, which detects the light partially transmitted by the second reflector 8. The light sensor 25 generates an electrical signal corresponding to the intensity of the detected light. This generated electrical signal can be transmitted to the controller 24 of the spectrometer 40, which will be described later, via, for example, a transmission cable connecting the laser processing device 30 and the spectrometer 40.

[0030] By detecting the position of the transmitted light as the optical sensor 25, for example, before the laser light 6 reaches the workpiece 70, the correlation between the signal intensity of the detection result and the output of the laser oscillator 1 can be obtained with high accuracy. However, the detection position is not particularly limited to this.

[0031] 1-3. Structure of Spectrometer Figure 3 This figure illustrates the structure of a spectroscopic device 40 according to this embodiment. The spectroscopic device 40 includes a collimating lens 15, a third reflector 16, a fourth reflector 17, a fifth reflector 18, condenser lenses 19, 20, and 21, a light sensor 22, a transmission cable 23, and a controller 24 within a housing 28. The housing 28 prevents stray light from entering the spectroscopic device 40 from outside and prevents light leakage from within.

[0032] Collimating lens 15 returns light transmitted from laser processing device 30 via optical fiber 13 to parallel light. Third reflector 16 transmits visible light with a wavelength of, for example, 400 nm to 700 nm, while reflecting all other components. Fourth reflector 17 reflects reflected light from laser beam 6 with a wavelength of, for example, approximately 1070 nm, while transmitting all other components. Fifth reflector 18 reflects thermal radiation with a wavelength of, for example, 1300 nm to 1550 nm.

[0033] The light that has passed through the collimating lens 15 is split into visible light, reflected light, and thermal radiation components by the third reflector 16, the fourth reflector 17, and the fifth reflector 18, and then focused by the condenser lenses 19 to 21. Furthermore, by placing arbitrary bandpass filters on the optical paths after the third reflector 16, the fourth reflector 17, and the fifth reflector 18, it is possible to select the wavelengths that are passed.

[0034] The optical sensor 22 includes, for example, optical sensors 22a, 22b, and 22c, each having high sensitivity to a different wavelength. The optical sensors 22a, 22b, and 22c detect the components of visible light, reflected light, and thermal radiation focused by the condenser lenses 19 to 21, respectively, and generate electrical signals corresponding to the intensity of the detected light. Alternatively, the optical sensor 22 may be comprised of a single optical sensor capable of detecting the intensity of each wavelength.

[0035] The electrical signal generated by the optical sensor 22 is transmitted to the controller 24 via a transmission cable 23. The controller 24 is a hardware controller that centrally controls the overall operation of the spectrometer 40. The controller 24 includes a CPU, communication circuitry, and other components, and transmits the electrical signal received from the optical sensor 22 to the inspection device 50. The controller 24 includes, for example, an A / D converter that converts the analog electrical signal into a digital signal (also referred to simply as a "signal"). To ensure a sufficient number of samples to capture the characteristics of the machining process and the trends of local values ​​of physical quantities, the sampling period for conversion to the digital signal is preferably, for example, less than 1 / 100 of the time during which the output of the laser 6 is controlled.

[0036] 1-4. Inspection device structure Figure 4 This is a block diagram illustrating the configuration of an inspection device 50 according to this embodiment. Inspection device 50 is comprised of, for example, an information processing device such as a computer. Inspection device 50 includes a CPU 51 for performing calculations, a communication circuit 52 for communicating with other devices, and a storage device 53 for storing data and computer programs.

[0037] The CPU 51 is an example of the arithmetic circuitry of the inspection device 50 in this embodiment. The CPU 51 executes a control program 56 stored in the storage device 53 to implement predetermined functions, including the construction of a determination model 57 and the inspection of the workpiece 70 using the constructed determination model 57. For example, the functions of the inspection device 50 in this embodiment are realized by the CPU 51 executing the control program 56. In this embodiment, the arithmetic circuitry of the inspection device 50, which is configured as the CPU 51, can be implemented using various processors such as an MPU or a GPU, or can be composed of one or more processors.

[0038] The communication circuit 52 is, for example, a communication circuit that communicates in accordance with standards such as IEEE 802.11, 4G, or 5G. The communication circuit 52 may also perform wired communication in accordance with standards such as Ethernet (registered trademark). The communication circuit 52 is capable of connecting to a communication network such as the Internet. Furthermore, the inspection device 50 can communicate directly with other devices via the communication circuit 52 or via an access point. Furthermore, the communication circuit 52 may be configured to communicate with other devices without going through a communication network. For example, the communication circuit 52 may include connection terminals such as a USB (registered trademark) terminal and an HDMI (registered trademark) terminal.

[0039] The storage device 53 is a storage medium that stores computer programs and data necessary to implement the functions of the inspection system 100. The storage device 53 stores a control program 56 executed by the CPU 51 and various data. After constructing a judgment model 57, the judgment model 57 is stored. The judgment model 57 is constructed through machine learning based on training data that correlates feature quantities representing the characteristics of the signals detected during laser processing under each condition with the surface roughness measured under each condition, for multiple processing conditions with varying surface roughness of the workpiece 70.

[0040] In this embodiment, the determination model 57 is a regression model implemented, for example, using linear regression, Lasso regression, ridge regression, decision tree, random forest, gradient boosting, support vector regression, Gaussian process regression, k-nearest neighbor method, or neural network. The determination model 57 of this embodiment outputs a numerical value representing the vertical displacement of the workpiece 70 from the reference surface as the surface roughness determination result. The construction of the determination model 57 will be described in detail later.

[0041] The storage device 53 is composed of, for example, a magnetic storage device such as a hard disk drive (HDD), an optical storage device such as an optical disk drive, or a semiconductor storage device such as a solid-state drive (SSD). The storage device 53 may include, for example, a temporary storage element composed of RAM such as DRAM or SRAM, or may function as internal memory of the CPU 51.

[0042] 2. Action In the inspection system 100 configured as described above, for example Figure 1 As shown, spectrometer 40 uses optical sensor 22 to detect the components of thermal radiation, visible light, and reflected light generated in melt portion 27 by irradiation with laser light 6. Spectrometer 40 transmits signals corresponding to the intensities of the detected components to inspection device 50. The following describes the operation of inspection device 50 in this system 100.

[0043] 2-1. Inspection and processing Below, use Figure 5 and Figure 6 The following describes an inspection process for inspecting the surface roughness when the laser processing apparatus 30 processes the workpiece 70 in the inspection apparatus 50 .

[0044] Figure 5 This is a flowchart illustrating the determination processing in the inspection device 50 according to this embodiment. Each process shown in this flowchart is executed, for example, by the CPU 51 of the inspection device 50. This flowchart is initiated, for example, when a user of the inspection system 100 or the like inputs a predetermined operation for starting the inspection process via an input device connected via the communication circuit 52.

[0045] First, the CPU 51 obtains, through the communication circuit 52 , signals corresponding to the components of thermal radiation, visible light, and reflected light detected by the optical sensor 22 of the spectroscopic device 40 ( S1 ).

[0046] Figure 6 It is a diagram for explaining signals acquired in the inspection device 50 . Figure 6 (A), (B), and (C) represent the signal waveforms corresponding to the intensities of thermal radiation, visible light, and reflected light, respectively. Figure 6 (D) represents the output of the laser beam 6 irradiated onto the workpiece 70 . Figure 6 The signals (A) to (C) correspond to the thermal radiation, visible light, and reflected light generated by the laser output. Figure 6 In (A) to (D), the horizontal axis represents time and the vertical axis represents signal strength ( Figure 6 (A)~(C)) or laser output ( Figure 6 (D)). In addition, Figure 6 In (A) to (D), the time interval T1 represents a time interval corresponding to one pulse of the laser light 6 , and the time interval T2 represents a time interval of peak output excluding the rise and fall of the laser output.

[0047] Here, in the laser processing device 30 of this embodiment, welding is performed on each workpiece 70 in a time interval T1 corresponding to one pulse of the laser beam 6. Figure 5 In step S1, as Figure 6 As shown in (A) to (C) of FIG. 8 , the CPU 51 obtains signals indicating changes in components of thermal radiation, visible light, and reflected light in a time interval T1 corresponding to the welding time of each workpiece 70 .

[0048] Next, the CPU 51 calculates a feature quantity (S2) to be input to the determination model 57 based on the acquired signal. The feature quantity is calculated, for example, based on the signal waveform representing the temporal variation of the signal intensity of each component, and includes an average intensity representing the average value of the signal intensity in time interval T2 and an integrated value of the signal intensity in time interval T2.

[0049] The CPU 51 inputs the feature values ​​calculated from the signals of each component detected during machining of the workpiece 70 into the determination model 57, and performs determination model processing (S3) to determine the surface roughness of the workpiece 70. In the determination model processing (S3) of this embodiment, the CPU 51 calculates a predicted value representing the surface roughness of the upper surface of the workpiece 70 irradiated with the laser beam 6. The relationship between these feature values ​​and surface roughness will be described in detail later.

[0050] The CPU 51 outputs the surface roughness value of the upper surface of the workpiece 70 calculated through the processing of the judgment model (S3) as the inspection result of the workpiece 70 (S4). For example, the CPU 51 may write the inspection result to the storage device 53 or transmit the inspection result to a location outside the inspection apparatus 50 via the communication circuit 52. The inspection result can be received and displayed by, for example, an information processing device or display device external to the inspection apparatus 50. Furthermore, the inspection apparatus 50 may include a display device (e.g., a monitor) capable of communicating with the CPU 51, so that the inspection result can be displayed on the display device.

[0051] Afterwards, CPU 51 ends Figure 5 Flowchart of the process. Figure 5 The flowchart is repeatedly executed each time welding processing is performed on each workpiece 70, for example.

[0052] Through the above inspection process, the inspection device 50 of this embodiment acquires the signal generated by the optical sensor 22 of the spectrometer 40 (S1), calculates a characteristic value based on the signal (S2), and then performs processing (S3) on the determination model 57 to inspect the surface roughness of the workpiece 70 based on the characteristic value. In this way, the surface roughness of the workpiece 70, the surface irradiated by the laser beam 6, can be inspected based on the signal of the light generated during laser processing, without directly measuring the surface roughness. This facilitates surface roughness inspection and, for example, allows the understanding of the impact of surface roughness fluctuations on the processing status during each processing step.

[0053] In addition, when the inspection device 50 described above is used at a manufacturing site of a product based on laser processing, for example, by setting a judgment standard on whether the surface roughness causes welding defects, defective welded products will not flow into subsequent processes, and defective welded products can be discharged based on the inspection results.

[0054] 2-2. Feature Quantities and Surface Roughness Regarding the relationship between the characteristic amount calculated based on the signal intensity of the light detected during laser processing in the above inspection process and the surface roughness, the following is used: Figure 7 To illustrate the findings obtained by the inventors of the technology in this disclosure.

[0055] Figure 7 This is a diagram for explaining the relationship between the feature amount calculated by the inspection device 50 and the surface roughness. Figure 7 (A) shows temporal changes in signal intensity of reflected light detected for each machining operation in each case where the surface roughness of the workpiece 70 is different. Figure 7 (B) is aimed at Figure 7 (A) Thermal radiation or visible light detected in the same situation, showing the temporal variation of signal intensity. Figure 7 (C) schematically shows the relationship between the surface roughness of the workpiece 70 and the molten zone 27 formed during the processing.

[0056] When the surface roughness of the workpiece 70 is different, the surface reflection of the laser beam 6 and / or the flow of the melt in the molten portion 27 in the workpiece 70 may change, thereby affecting the shape of the molten portion 27. Figure 7 As shown in (C), when the surface roughness changes, the shape of the molten portion 27 changes, and the amount of luminous and scattered light in the molten portion 27 changes accordingly, as shown in FIG. Figure 7 As shown in (A) and (B), the detected signal intensity changes.

[0057] If the surface roughness is low, for example, the molten metal formed by laser light 6 in the molten zone 27 is less likely to spread in the molten zone width direction, i.e., in a direction perpendicular to the scanning direction. This allows the input heat to be concentrated while maintaining the shape. Therefore, if the melting temperature rises, the surface temperature of the molten zone 27 increases, and the luminescence amount and corresponding signal intensity are relatively large. On the other hand, if the surface roughness is high, the molten metal in the molten zone 27 is more likely to spread in the molten zone width direction, and the heat may be dispersed. Therefore, if the surface temperature of the molten zone 27 decreases, the luminescence amount and corresponding signal intensity are relatively small.

[0058] Based on the above findings, the inspection device 50 of this embodiment constructs a determination model 57 for determining the surface roughness of the workpiece 70 using a feature value corresponding to the signal intensity, based on a signal corresponding to at least one of the components of thermal radiation, visible light, and reflected light during machining, through a training process described later. The feature values ​​input into this determination model 57 are described below.

[0059] For example, the CPU 51 of the inspection device 50 calculates the average intensity of each signal as a feature value in a time interval T2 corresponding to the peak output period of each processing of the laser oscillator 1 of the laser processing device 30. The time interval T2 can be determined based on the output waveform of the laser oscillator 1, for example.

[0060] Furthermore, as described above, when the surface roughness of the workpiece 70 increases or decreases, in addition to affecting the shape of the molten zone 27, the temperature of the portion of the workpiece 70 irradiated with the laser beam 6 also changes, causing changes in the amount of reflected light, thermal radiation, and visible light from the molten zone 27. In the inspection apparatus 50 of this embodiment, the CPU 51 calculates, as a characteristic quantity corresponding to such changes in light intensity, the integrated value of the signal intensity during the time interval T2 when the laser beam 6 is at its peak output, in addition to the average intensity.

[0061] 2-3. Training of the judgment model Below, use Figure 8 and Figure 9 The training process for constructing the determination model 57 will be described.

[0062] Figure 8 This is a flowchart illustrating a training process of the determination model 57 used in the inspection process. Each process in this flowchart is executed by, for example, the CPU 51 of the inspection device 50 .

[0063] First, the CPU 51 acquires training data stored in advance in the storage device 53 , for example ( S11 ).

[0064] Figure 9 This diagram illustrates training data D1 for the determination model 57. Training data D1 is data that associates characteristic quantities for each machining operation of the workpiece 70 with, for example, actual surface roughness values ​​of the workpiece 70 measured before the machining operation. Training data D1 is constructed by, for example, performing laser machining using the laser machining device 30 under a plurality of varying welding machining conditions, acquiring data such as signals detected by the spectrometer 40, and actual surface roughness values.

[0065] Figure 9In addition to the characteristic values ​​of the average intensity and integral value calculated based on the signals of each component of thermal radiation, visible light, and reflected light, the training data D1 also records the output of the laser oscillator 1 in association with each condition. As the measured value of the surface roughness, for example, based on the measurement results of the upper surface of the workpiece 70 using a shape measuring instrument, the arithmetic mean height Ra or maximum height Rz of the line roughness representing the two-dimensional surface properties, or the arithmetic mean height Sa or maximum height Sz of the surface roughness representing the three-dimensional surface properties is calculated. The surface roughness can also include multiple such indicators. In addition, multiple measurements and laser processing can be performed under each condition, and the average value of the multiple data obtained can be used as the measured value and characteristic value of the condition.

[0066] As a measure of surface roughness, for example, the surface of the workpiece 70 can be polished using sandpaper of different grits to vary the height displacement from a predetermined reference surface. Furthermore, multiple surface roughness conditions can be set for each grit of sandpaper, but this is not a limitation. Furthermore, in laser processing for overlap welding, joint strength measurements can be obtained after processing under various conditions to ensure that the desired joint strength of the workpiece 70 after processing is correlated with the surface roughness of the workpiece 70. Details of the generation of training data D1 will be described later.

[0067] return Figure 8 The CPU 51 performs machine learning using the acquired training data D1 to generate a determination model 57 by calculating the corresponding surface roughness based on the feature quantity (S12). In step S12, the CPU 51 performs machine learning on the determination model 57 to minimize the error between the surface roughness determined by the determination model 57 based on the feature quantity for each condition in the training data D1 and the surface roughness in the training data D1 for each condition.

[0068] Through the above training process, the determination model 57 can be generated as a learned model that calculates a predicted value of surface roughness based on feature quantities calculated from signals of each component of thermal radiation, visible light, and reflected light detected during laser processing.

[0069] In addition, the training process of the determination model 57 may be performed in an information processing device different from the inspection device 50. The inspection device 50 may also obtain the determination model constructed by the communication circuit 52 via a communication network. Figure 9 For example, it may include feature quantities calculated for some components of thermal radiation, visible light, and reflected light, or it may include only one of the average intensity and the integrated value.

[0070] 2-4. Training Data Generation Process The inspection device 50 of this embodiment, for example, performs a process of generating training data D1 for the determination model 57 before the above-mentioned process of generating the determination model 57. This process includes, for example, pre-processing for generating the determination model 57 with high accuracy based on the training data D1 obtained by this process. Figure 10 The generation process of such training data D1 will be described.

[0071] Figure 10 This is a flowchart illustrating the process of generating training data D1. For example, the process in this flowchart begins with the surface roughness of a workpiece 70, similar to the inspection target, being varied under various conditions, with the surface roughness measurement results obtained before processing and signals of various components, such as reflected light, detected during laser processing. The surface roughness measurement results and signals under these various processing conditions are stored, for example, in the storage device 53 of the inspection device 50. The various processing conditions are, for example, pre-set and stored in the storage device 53. Furthermore, each process in this flowchart is executed, for example, by the CPU 51 of the inspection device 50.

[0072] Furthermore, in this embodiment, to generate training data D1 capable of constructing a more accurate determination model 57, predetermined preprocessing (S23, S25-S27, etc.) is performed on the surface roughness measurement results and detected signals before generating training data D1. For example, signals during machining vary not only due to changes in the surface roughness of the workpiece 70 but also due to foreign matter such as dirt adhering to the surface. If training data D1 contains a large number of signals that vary due to factors other than surface roughness, the determination model 57 trained using this training data D1 may have difficulty learning the correlation between the signal's feature value and surface roughness. The preprocessing in this embodiment includes processing to suppress the inclusion of such signals in the training data D1.

[0073] exist Figure 10 In the processing, the CPU 51 first obtains the surface roughness measurement results obtained by a profilometer or the like under a specific processing condition from the storage device 53 (S21). The accuracy of the profilometer is preferably capable of measuring to the nanometer (nm) level, but there is no particular limitation as long as the accuracy is in the micrometer (μm) level. The area on the surface of the workpiece 70 to be measured can be determined based on the weld shape, for example, the area corresponding to the range of the irradiation laser 6. Workpieces 70 with varying surface roughness can be produced, for example, by varying the number of sandpaper used to polish the surface of the workpiece 70 before processing by every 100.

[0074] In addition to measuring surface roughness, the joint strength of the workpiece 70 after welding can also be measured. For example, tensile strength or torque strength can be measured using a tensile testing machine, and the measurement method is not particularly limited. In this case, the joint strength measurement results can be used together with the training data D1 to manage whether the desired joint strength is maintained, for example, in association with each condition.

[0075] Next, the CPU 51 calculates the arithmetic mean height and maximum height based on the surface roughness measurement results (S22). For example, the arithmetic mean height Ra and maximum height Rz of the line roughness, or the arithmetic mean height Sa and maximum height Sz of the surface roughness, are calculated based on the measurement method. The surface roughness parameters Sa and Sz are three-dimensionally expanded from the line roughness parameters Ra and Rz calculated from the profile curve based on the measurement results. For example, the arithmetic mean heights Ra and Sa are calculated using the following calculations.

[0076] [Formula 1] [Formula 2] Here, b represents a reference length of a profile curve in measuring line roughness, A represents a reference area in measuring surface roughness, and Z represents a coordinate value in the height direction based on the measurement result.

[0077] The following description uses an example in which the arithmetic mean height Sa and maximum height Sz of the surface roughness are calculated in step S22. The arithmetic mean height is sometimes referred to simply as the "average height." The maximum height Sz is calculated as the sum of the maximum peak height and the maximum valley depth in the reference area A.

[0078] The CPU 51 determines whether the calculated maximum height Sz is less than or equal to the average height Sa multiplied by a predetermined value, that is, whether the maximum height Sz and the average height Sa satisfy the relationship "Sz ≤ Sa × predetermined value" (S23). When measuring surface roughness, if foreign matter, such as foreign matter, adheres to the measurement area on the surface of the workpiece 70, the measured value may vary from the actual surface roughness. The type of foreign matter may include, for example, resin materials or abrasive particles, and is not particularly limited.

[0079] When the maximum height Sz and the average height Sa are in the relationship of “Sz≦Sa×predetermined value” ( S23 : YES), the CPU 51 obtains the signal detected by the optical sensor 22 during machining for the machining conditions for which the measurement results used for calculating the parameters Sz and Sa were obtained ( S24 ).

[0080] On the other hand, if the relationship "Sz ≤ Sa × specified value" is not met, that is, if the maximum height Sz is greater than the value obtained by multiplying the average height Sa by the specified value (S23: No), the CPU 51 skips step S24 and proceeds to step S30. For example, the CPU 51 sets the processing condition used to calculate the parameters Sz and Sa as the target for re-measurement, or shifts the processing condition to a condition that, among multiple pre-set processing conditions, allows processing to be performed following the pre-set condition (S30).

[0081] For example, the CPU 51 determines whether a re-measurement or next processing condition exists among the plurality of processing conditions ( S31 ). If a re-measurement or next processing condition exists ( S31 : Yes), the CPU 51 returns to step S21 . The CPU 51 obtains the re-measured surface roughness measurement result or the measurement result under the next processing condition ( S21 ) and repeats the processing from step S22 onward.

[0082] Through the above-described processing, using specified values ​​and based on the relationship between maximum height Sz and average height Sa (S23), signals under machining conditions for each parameter Sz and Sa are obtained (S24), or re-measured (S30). This eliminates deviations in measurement results caused by interference factors such as foreign matter, allowing the results of highly accurate surface roughness measurements under machining conditions to be correlated with signals detected during machining under those conditions, thereby constructing training data D1.

[0083] The specified values ​​of step S23 can also be experimentally set, taking into account the fluctuations of the parameters Sz and Sa caused by the aforementioned interference factors. Furthermore, the calculation used to determine the relationship between the parameters Sz and Sa is not limited to the example of step S23. For example, in addition to measuring surface roughness, when foreign matter or dirt is detected on the surface of the workpiece 70 captured by a camera through image processing, step S30 can be performed so that the measurement result is excluded from the training data D1. Furthermore, after step S30, when re-measuring the surface roughness, the surface roughness near the molten zone 27 after welding can be measured. However, it is also possible to measure the surface roughness of the same workpiece 70 as before the re-measurement and the area welded under the processing conditions, then re-weld and perform the subsequent processing. It is assumed that the surface roughness of the molten zone 27 and its vicinity is approximately the same. The surface roughness measurement location is not particularly limited, as long as it is the surface of the workpiece 70 irradiated with the laser beam 6. However, the area before welding where the molten zone 27 is formed is more preferred.

[0084] After obtaining the signal (S24), the CPU 51 may also perform a process (S25) to correct the start time of the signal, for example, to make the rise start time (i.e., the start time) of the signal waveforms of the multiple signals obtained each time step S24 is executed uniform. Figure 6 The start times of the signals shown in (A) to (C) can be set by the inspection device 50 receiving a trigger signal output from the laser processing device 30 when the laser beam 6 oscillates, and the trigger signal is received at the time of receipt. However, errors may occur in the start times due to, for example, errors in the start time of the trigger signal.

[0085] In step S25, the CPU 51 corrects the signal start time by shifting it, for example, based on the time it takes for the signal strength to reach a predetermined value (e.g., 0.2 V) during the rising edge of the signal waveform. This correction allows the conditions for calculating the feature value (S26), described later, to be standardized across all signals, resulting in training data D1 that can be used to more accurately construct the determination model 57.

[0086] Next, the CPU 51 sets, for example, the time interval T2 corresponding to the peak output of the laser output as a predetermined time interval in the signal for each processing operation, and calculates the proportion of periods during which the signal intensity in time interval T2 exceeds a predetermined threshold value based on the acquired signal (S26). The predetermined threshold value is pre-set based on the average signal waveform, i.e., the average waveform, for each component of reflected light, thermal radiation, and visible light, and stored in the storage device 53. The average waveform is calculated by, for example, averaging the signal intensities of the signals acquired during multiple laser processing operations under each processing condition. For example, the value obtained by adding the standard deviation of the signal intensities to the average value of the signal intensities in time interval T2 of the average waveform is set as the upper threshold value, and the value obtained by subtracting the average value is set as the lower threshold value.

[0087] In step S26, the CPU 51 calculates the ratio (also referred to as the "NG ratio") of the signal strength of the acquired signal exceeding the threshold value of the average waveform described above (i.e., exceeding the upper threshold value or falling below the lower threshold value). The NG ratio can also be calculated as a percentage by multiplying the calculated value based on the following calculation formula (1) by "100." In the following calculation formula (1), the number of sampling points of the corresponding signal can be used as the time interval T2 and the period exceeding the threshold value.

[0088] NG ratio = period exceeding the threshold value in time interval T2 / time interval T2 (1) After calculating the NG ratio of the acquired signal ( S26 ), the CPU 51 determines whether the calculated NG ratio is smaller than a predetermined value (eg, 20%) ( S27 ).

[0089] If the calculated NG ratio is less than the specified value (S27: Yes), the CPU 51 calculates the feature value based on the acquired signal (S28). For example, as described above, the CPU 51 calculates the average value and the integral value of the signal strength in the time interval T2 as the feature value. In step S28, the CPU 51 may also calculate the feature value based on the value obtained in the inspection process ( Figure 5 ) to calculate other feature quantities.

[0090] On the other hand, if the calculated NG ratio is greater than the specified value (S27: No), the CPU 51 does not specifically calculate the characteristic value (S28). Similarly to the case where the average height Sa and the maximum height Sz of the surface roughness do not have the specified relationship (S23: No), the process proceeds to step S30. For example, the CPU 51 sets the processing condition for which the ratio was calculated as a target for re-measurement, or shifts the processing condition to the next processing condition (S30).

[0091] According to the above processing, based on the proportion of signal strength exceeding the threshold value of the average waveform within a predetermined time interval, a feature value is calculated based on the signal strength (S27), or (S28), or re-measurement is performed (S30). In this way, when generating training data D1, it is possible to exclude signals with abnormal waveforms caused by, for example, the generation of interference, while calculating and using feature values ​​based on signals that are expected to be less affected by interference.

[0092] The CPU 51 associates the feature value calculated in step S28 with the calculated surface roughness value calculated based on the measurement results under the processing conditions for calculating the feature value, and the calculated surface roughness value with the parameters Sa and Sz being in the prescribed relationship in step S23, and adds the result to the training data D1 (S29). Figure 5 The inspection result outputted in the above example may be either the average height Sa or the maximum height Sz, or both of them may be added. The CPU 51 and the storage device 53 store or update the added training data D1.

[0093] When the feature value and the calculated value of the surface roughness are added to the training data D1 ( S29 ), the CPU 51 proceeds to step S31 . If the next processing condition exists among the multiple processing conditions ( S31 : Yes), the processing from S21 onwards is repeated for the next processing condition.

[0094] On the other hand, if there is no re-measurement or next processing condition (S31: No), the CPU 51 ends the processing of this flowchart.

[0095] According to the above processing, for each processing condition, the calculated value calculated based on the surface roughness measurement result (S21, S22) and the feature value calculated based on the signal such as the reflected light detected during processing (S24, S28) are associated and added to the training data D1 (S29). Using the training data D1 generated by such generation processing, the training processing of the judgment model 57 can be performed ( Figure 8 ). Furthermore, in the process of generating training data D1 in this embodiment, pre-processing is performed on the acquired surface roughness and signals by eliminating data with a high probability of variation due to interference, etc. (S23, S26-S27), and correcting the start time of the unified signal (S25). This, for example, can improve the quality of the data included in training data D1.

[0096] The threshold value for the average waveform in step S26 can be set based on the signal after the start time has been corrected using the same process as in step S25, or based on the average value and standard deviation of the signal obtained in step S24. Alternatively, a value obtained by multiplying the standard deviation by a predetermined ratio can be used in place of the standard deviation. Furthermore, the period for calculating the percentage exceeding the threshold in step S26 is not limited to time interval T2; for example, it can be set by the user of the inspection apparatus 50 via the communication circuit 52 or the like.

[0097] The predetermined value in step S27 is preferably changeable by the user, but may be automatically set by the inspection device 50. The predetermined value may be set for the laser output, preferably for each of reflected light, thermal radiation, and visible light, but may also be set to a common value.

[0098] 3. Effects, etc. As described above, the inspection process (S1 to S4) of the present embodiment provides an inspection method for a workpiece 70 during laser processing. The method includes: a step (S1) of obtaining a signal representing a change in the component in a time interval T1 corresponding to the welding time of each workpiece 70, the signal being generated by the optical sensor 22 detecting at least one of the components of thermal radiation, visible light, and reflected light generated by irradiating the workpiece 70 with the laser 6; a step (S2) of calculating a feature quantity representing a feature of the signal in a time interval T2 (an example of a predetermined interval) in the time interval T1; a step (S3) of inputting the calculated feature quantity into a determination model 57 for determining the surface roughness representing the surface properties of the workpiece 70 irradiated with the laser 6, and determining the surface roughness of the workpiece 70; and a step (S4) of outputting the determined surface roughness as an inspection result. The determination model 57 is constructed based on training data D1 that includes a feature quantity calculated from a signal of a component detected by laser processing under each of a plurality of conditions that change the surface roughness, and a correlation with the surface roughness of each condition (see Figure 8 and Figure 9 ).

[0099] According to the above method, a signal is generated by detecting one or more components of thermal radiation, visible light, and reflected light generated by irradiating the workpiece 70 with laser light 6 (S1), and a feature value is calculated based on the signal (S2). Then, a predicted value of surface roughness is calculated based on the feature value using the judgment model 57 constructed based on the training data D1 (S3). Using the judgment model 57, for example, surface roughness inspection can be performed based on the predicted value, even without directly measuring the surface roughness, making surface roughness inspection easier.

[0100] In this embodiment, the surface roughness of the training data D1 is measured in an area corresponding to the range of the laser beam 6 irradiated onto the surface of the workpiece 70 under various conditions (similar to the inspection object). For example, in laser processing for overlap welding, the surface roughness can also be measured in an area of ​​the surface of the upper member of the workpiece 70 on the side irradiated with the laser beam 6 where the molten zone 27 is formed by welding.

[0101] In this embodiment, the time interval T2 corresponds to the period during which each workpiece 70 is irradiated with the laser beam 6 at the peak output (see Figure 6 ), the feature quantity includes the average intensity of the signal in the time interval T2. Figure 7 As illustrated, the signal intensity at the peak output may vary depending on the surface roughness. Therefore, it is considered that the surface roughness can be predicted with high accuracy by using the average intensity of the time interval T2 as the feature value.

[0102] In this embodiment, the feature value also includes the integral value of the signal in the time interval T2. Figure 7 As described above, surface roughness may affect the shape of the molten portion 27 formed during processing. Therefore, the amount of heat input due to irradiation with the laser beam 6 may change, and the amount of light from the surface may increase or decrease along with the change in the surface temperature of the molten portion 27. When the integral value is used, it is believed that such a change in the amount of light can be reflected in the characteristic value, and the surface roughness can be predicted with high accuracy.

[0103] In this embodiment, surface roughness includes an arithmetic mean height Sa and a maximum height Sz calculated based on vertical displacement from a reference surface (an example of a reference plane) of the workpiece 70. This method further includes a step (S23) of determining whether, before constructing the determination model 57 based on the training data D1, the arithmetic mean height Sa and the maximum height Sz calculated from surface roughness measurements for each condition of the workpiece 70 satisfy a relationship of "maximum height Sz ≤ arithmetic mean height Sa × a predetermined value," as an example of a predetermined relationship; and a step (S23: Yes, S24, S29) of selectively generating training data D1 by including the surface roughness and feature values ​​for conditions satisfying this relationship among a plurality of conditions. This allows the generation of training data D1 while suppressing the influence of interference factors, such as variations in surface roughness measurement results caused by foreign matter adhering to the surface of the workpiece 70. Based on this training data D1, a highly accurate determination model 57 can be constructed.

[0104] The method in this embodiment further includes a step (S26) of calculating, before constructing a determination model based on training data D1, the proportion of intervals in time interval T2 (an example of a predetermined interval) in which the acquired signal intensity exceeds the threshold value based on an average waveform threshold value (an example of a threshold value set for the signal intensity of a component detected under each of a plurality of conditions); and (S27-S29) of generating training data D1 by selectively calculating and including feature quantities from a plurality of conditions by comparing the calculated NG ratio with a predetermined value (an example of a predetermined ratio). For example, when various interference factors cause fluctuations in signal intensity, an abnormal signal waveform may occur. The above process allows, even in such cases, for example, if the proportion of signal intensity exceeding the threshold value exceeds the predetermined ratio, to identify an abnormal signal waveform, and training data D1 can be selectively generated from a plurality of conditions so that feature quantities derived from the signal are not included in training data D1. This also suppresses the influence of interference factors in training data D1.

[0105] In this embodiment, the determination model 57 is generated by machine learning in a manner that minimizes the error between the surface roughness determined based on the characteristic values ​​of each condition in the training data D1 and the surface roughness of each condition in the training data D1 (S11, S12). Through machine learning using the training data D1, which thus associates the characteristic values ​​of each condition with the surface roughness measured under that condition, the determination model 57 for determining the surface roughness of the workpiece 70 is obtained based on the characteristic values ​​calculated based on the signals detected during the machining of the workpiece 70.

[0106] In this embodiment, the surface roughness includes, as an example of a numerical value representing a vertical displacement from a reference plane of the surface of the workpiece 70, an arithmetic mean height Sa or Ra and / or a maximum height Sz or Rz. The surface roughness determined by the determination model 57 is not limited thereto and may include other parameters of surface roughness or line roughness.

[0107] In the inspection system 100 of this embodiment, the inspection device 50 is an example of an inspection device for a workpiece 70 being laser-processed. The inspection device 50 includes a CPU 51, an example of a calculation circuit, and a communication circuit 52. The communication circuit 52 receives a signal generated by the optical sensor 22 detecting at least one of the components of thermal radiation, visible light, and reflected light generated by irradiating the workpiece 70 with the laser beam 6. The signal represents the change in the component during a time interval T1, as an example of a time interval corresponding to the welding time of each workpiece 70. The CPU 51 receives the signal through the communication circuit 52 (S1), calculates a feature quantity representing the characteristics of the signal during a time interval T2 (an example of a predetermined interval) within time interval T1 (S2), inputs the calculated feature quantity into a surface roughness determination model 57 that determines the surface properties of the workpiece 70 irradiated with the laser beam 6, determines the surface roughness of the workpiece 70 (S3), and outputs the determined surface roughness as an inspection result (S4). The determination model 57 is constructed based on training data D1 including feature quantities calculated from signals of components detected by laser processing under each of a plurality of conditions that vary the surface roughness, in association with the surface roughness under each condition.

[0108] According to the above-described inspection apparatus 50 , the surface roughness of the workpiece 70 can be easily inspected by executing the above-described inspection method.

[0109] (Other embodiments) As described above, the above embodiments are described as examples of the technology disclosed in this application. However, the technology disclosed in this application is not limited thereto and can also be applied to embodiments that have been appropriately modified, replaced, added, or omitted. In addition, it is also possible to combine the various components described in the above embodiments to form new embodiments.

[0110] In the above-mentioned first embodiment, the generation process of the training data D1 ( Figure 10 ) to obtain the measurement results of the surface roughness of the workpiece 70 and the signals detected during machining (S21, S24). In this embodiment, in addition to this, for example, the measurement results of the width, length, and / or area of ​​the molten zone 27 in the workpiece 70 after welding and / or the measurement results of the penetration depth can also be obtained. For example, from the perspective of improving the accuracy of effectively utilizing the surface roughness prediction, such additional measurement results can also be used as feature quantities of the judgment model 57.

[0111] In the above embodiments, examples were described in which signals detected during machining under various machining conditions were acquired during the generation of training data D1 (S24). In this embodiment, multiple machining operations under various machining conditions can be performed multiple times, and multiple signals detected during each machining operation can be acquired in step S24. In this case, for example, after correcting the start time of each signal, similar to step S25, subsequent processing can be performed based on the average waveform of the multiple signals. This can, for example, reduce the influence of fluctuations in the signal waveform caused by interference during the detection of light from each signal on the feature values ​​calculated from the signal.

[0112] In the above-mentioned embodiments, an example is described in which the NG ratio is calculated as the ratio of the signal intensity exceeding the threshold value in the time interval T2 during the generation process of the training data D1 (S26). In this embodiment, the calculation is not limited to the time interval T2, and for example, the calculation may be performed in the time interval T1 corresponding to one pulse of the laser light 6 (see Figure 6 ), or it can be calculated during the acquisition period corresponding to one waveform of the signal without setting a special interval.

[0113] In the above embodiments, examples have been described in which preprocessing of the surface roughness and signals obtained by the inspection device 50 is performed during the generation of training data D1 ( S23 , S25 - S27 ). While such preprocessing is preferred for generating training data D1 for constructing a more accurate determination model 57 , in this embodiment, it is not specifically performed and can be performed at the user's discretion.

[0114] In the above embodiments, examples have been described in which the training data D1 generation process is executed in the inspection device 50. In this embodiment, the training data D1 generation process is not limited to being executed by the inspection device 50, but may be executed by an external information processing device.

[0115] While the aforementioned embodiments illustrate overlap welding as an example of laser processing within this disclosure, the present disclosure can be applied to determining the surface roughness of workpieces in various welding processes. Furthermore, as laser processing other than welding, the present disclosure can also be applied to laser cutting or drilling. In such processes, for example, by constructing a surface roughness determination model for the surface of the workpiece irradiated with laser light during processing, as in the aforementioned embodiments, the surface roughness of the processed area can be easily inspected.

[0116] The present disclosure is not limited to the above-described embodiments, and various modifications are possible. In other words, embodiments obtained by combining technical means appropriately modified by those skilled in the art also fall within the scope of the present disclosure.

[0117] (Method of this disclosure) As described above, the present disclosure includes the following aspects.

[0118] (First Method) A method of checking, This is a method for inspecting the workpiece during laser processing, including: a step of acquiring a signal generated by detecting, using a light sensor, at least one of the components of thermal radiation, visible light, and reflected light generated by irradiating the workpiece with a laser, and indicating a change in the component in a time interval corresponding to a processing time of each of the workpieces; a step of calculating a feature quantity representing a feature of the signal in a predetermined interval within the time interval; a step of inputting the calculated feature value into a surface roughness determination model for determining the surface properties of the surface of the workpiece irradiated with the laser beam, thereby determining the surface roughness of the workpiece; and The step of outputting the calculated predicted value of the surface roughness as an inspection result, The determination model is constructed based on training data including a feature amount calculated from a signal of the component detected by performing the laser processing under each of a plurality of conditions that vary the surface roughness, in association with the surface roughness under each condition.

[0119] (Second Method) According to the inspection method described in the first aspect, The surface roughness of the training data is measured in a region corresponding to a range where the surface of the workpiece is irradiated with the laser light under each of the conditions.

[0120] (Third Method) The inspection method according to the first or second aspect, wherein: The predetermined interval corresponds to a period during which each of the workpieces is irradiated with the laser beam at peak output. The feature value includes an average intensity of the signal in the predetermined interval.

[0121] (Fourth Method) The inspection method according to any one of the first to third aspects, wherein: The predetermined interval corresponds to a period during which each of the workpieces is irradiated with the laser beam at peak output. The feature amount includes an integrated value of the signal in the predetermined interval.

[0122] (Fifth Method) The inspection method according to any one of the first to fourth aspects, wherein: The surface roughness includes an arithmetic mean height and a maximum height calculated from a vertical displacement from a reference plane of the surface of the workpiece. The inspection method further comprises: Before constructing the determination model based on the training data, a step of determining whether the arithmetic mean height and the maximum height calculated by measuring the surface roughness of the workpiece under each condition are in a predetermined relationship; and A step of generating the training data by selectively including the arithmetic mean height and the maximum height calculated as the surface roughness of the condition in the predetermined relationship among the plurality of conditions and the feature amount under the condition.

[0123] (Sixth Method) The inspection method according to any one of the first to fifth aspects, wherein: The inspection method further comprises: Before constructing the decision model based on the training data, a step of calculating, for the signal of the component detected under each of the plurality of conditions, based on a threshold value set for signal intensity, a ratio of intervals in which the intensity of the signal exceeds the threshold value in the predetermined interval; and The step of generating the training data by comparing the calculated ratio with a predetermined ratio so as to selectively calculate and include the feature amount from among the plurality of conditions.

[0124] (Seventh Method) An inspection method according to any one of the first to sixth aspects, wherein the determination model is generated by machine learning in a manner that minimizes the error between the surface roughness determined based on the feature quantity of each condition in the training data and the surface roughness of each condition in the training data.

[0125] (Eighth Method) The inspection method according to any one of the first to seventh aspects, wherein the surface roughness includes a numerical value indicating a displacement in a vertical direction from a reference plane of the surface of the workpiece.

[0126] (Ninth Method) An inspection device, This is a device for inspecting workpieces during laser processing, equipped with: an operational circuit; and a communication circuit receiving a signal generated by detecting, by a light sensor, at least one of the components of thermal radiation, visible light, and reflected light generated by irradiating the workpiece with laser light; The signal is a signal indicating a change in the component in a time interval corresponding to a processing time of each of the workpieces. The arithmetic circuit performs: Obtaining the signal through the communication circuit, In a predetermined interval of the time interval, a feature quantity representing a feature of the signal is calculated, The calculated feature value is input into a surface roughness determination model for determining the surface properties of the surface of the workpiece irradiated with the laser beam, thereby determining the surface roughness of the workpiece. The calculated predicted value of the surface roughness is output as an inspection result, The determination model is constructed based on training data including a feature amount calculated from a signal of the component detected by performing the laser processing under each of a plurality of conditions that vary the surface roughness, and the surface roughness under each condition in association with the feature amount.

[0127] (Tenth Method) The inspection device according to the ninth aspect, wherein The determination model is generated by machine learning so as to minimize an error between the surface roughness determined based on the feature amount of each condition in the training data and the surface roughness of each condition in the training data.

[0128] [Industrial Applicability] The present disclosure can be applied to a method and apparatus for inspecting a workpiece for determining the surface roughness of the workpiece irradiated with laser light in various laser processes such as overlap welding.

[0129] Explanation of symbols 1 Laser oscillator 2 Optical fiber for laser transmission 3 Lens barrel 4 Collimating lens 5.11 Condenser lens 6 Laser 7 First reflector 8 Second reflector 13 Fiber Optic 15 Collimating lens 16 Third reflector 17 Fourth Reflector 18 Fifth Reflector 19, 20, 21 Condenser lens 22 Light Sensor 23 Transmission Cable 24 Controller 25 Light Sensor 26 Pressing fixture 27 Melting area 30 Laser processing equipment 40 Spectrometer 50 Inspection device 51 CPU 52 Communication Circuit 53 Storage Device 56 Control Program 57 Judgment Model 70 Workpiece D1 training data 100 Inspection System

Claims

1. A method of inspection, This is a method for inspecting the workpiece during laser processing, including: a step of acquiring a signal generated by detecting, using a light sensor, at least one of the components of thermal radiation, visible light, and reflected light generated by irradiating the workpiece with a laser, and indicating a change in the component in a time interval corresponding to a processing time of each of the workpieces; a step of calculating a feature quantity representing a feature of the signal in a predetermined interval within the time interval; a step of inputting the calculated feature value into a surface roughness determination model for determining the surface roughness of the workpiece irradiated with the laser beam, thereby determining the surface roughness of the workpiece; as well as The step of outputting the calculated predicted value of the surface roughness as an inspection result, The determination model is constructed based on training data including a feature amount calculated from a signal of the component detected by performing the laser processing under each of a plurality of conditions that vary the surface roughness, in association with the surface roughness under each condition.

2. The inspection method according to claim 1, wherein: The surface roughness of the training data is measured in a region corresponding to a range where the surface of the workpiece is irradiated with the laser light under each of the conditions.

3. The inspection method according to claim 1, wherein: The predetermined interval corresponds to a period during which each of the workpieces is irradiated with the laser beam at peak output. The feature value includes an average intensity of the signal in the predetermined interval.

4. The inspection method according to claim 1, wherein: The predetermined interval corresponds to a period during which each of the workpieces is irradiated with the laser beam at peak output. The feature amount includes an integrated value of the signal in the predetermined interval.

5. The inspection method according to claim 1, wherein: The surface roughness includes an arithmetic mean height and a maximum height calculated based on a vertical displacement from a reference plane of the surface of the workpiece. The inspection method further comprises: Before constructing the decision model based on the training data, a step of determining whether the arithmetic mean height and the maximum height calculated by measuring the surface roughness of the workpiece under the above-mentioned conditions are in a predetermined relationship; as well as A step of generating the training data by selectively including the arithmetic mean height and the maximum height calculated as the surface roughness of the condition in the predetermined relationship among the plurality of conditions and the feature amount under the condition.

6. The inspection method according to claim 1, further comprising: Before constructing the decision model based on the training data, a step of calculating, for the signal of the component detected under each of the plurality of conditions, based on a threshold value set for signal intensity, a ratio of intervals in which the intensity of the signal exceeds the threshold value in the predetermined interval; and The step of generating the training data by comparing the calculated ratio with a predetermined ratio so as to selectively calculate and include the feature amount from among the plurality of conditions.

7. The inspection method according to claim 1, wherein: The determination model is generated by machine learning so as to minimize an error between the surface roughness determined based on the feature amount of each condition in the training data and the surface roughness of each condition in the training data.

8. The inspection method according to claim 1, wherein: The surface roughness includes a numerical value indicating a displacement in a vertical direction from a reference plane of the surface of the workpiece.

9. An inspection device, This is a device for inspecting workpieces during laser processing, equipped with: an operational circuit; and a communication circuit receiving a signal generated by detecting, by a light sensor, at least one of the components of thermal radiation, visible light, and reflected light generated by irradiating the workpiece with laser light; The signal is a signal indicating a change in the component in a time interval corresponding to a processing time of each of the workpieces. The arithmetic circuit performs: Obtaining the signal through the communication circuit, In a predetermined interval within the time interval, a feature quantity representing a feature of the signal is calculated, The calculated feature value is input into a determination model for determining the surface roughness of the surface of the workpiece irradiated with the laser beam, thereby determining the surface roughness of the workpiece. The calculated predicted value of the surface roughness is output as an inspection result, The determination model is constructed based on training data including a feature amount calculated from a signal of the component detected by performing the laser processing under each of a plurality of conditions that vary the surface roughness, and the surface roughness under each condition in association with the feature amount.

10. The inspection device according to claim 9, wherein: The determination model is generated by machine learning so as to minimize an error between the surface roughness determined based on the feature amount of each condition in the training data and the surface roughness of each condition in the training data.

Images