JP7612697B2 - Soldering System - Google Patents

Description

The present disclosure relates to a soldering system.

Soldering systems equipped with a means for automatically checking the quality of soldering are known. For example, the system in Patent Document 1 uses a non-contact temperature sensor to detect the temperature distribution of a board immediately after soldering, performs image processing, and detects the presence or absence of an abnormality in the soldering temperature by comparing the processed signal with optimal temperature data previously recorded in a computer. Based on the detection results, this system controls the board transport speed, the preheater temperature, and the temperature of the molten solder.

Japanese Unexamined Patent Publication No. 7-142852

In Patent Document 1, it is not possible to continuously monitor changes in the contact state between the board and the jetted molten solder. As a result, the quality of the soldering on the board is not stable.

Therefore, the object of the present disclosure is to provide a soldering system capable of soldering circuit boards with stable quality.

The soldering system disclosed herein includes a flux applicator that applies flux to a substrate, a preheater that preheats the substrate, a solder bath that stores molten solder, a solder bath heater that melts the solder in the solder bath, a jet nozzle that jets the molten solder in the solder bath toward the substrate, a transport mechanism that transports the substrate sequentially above the flux applicator, above the preheater, and above the solder bath, and a temperature measuring device positioned above the jet nozzle.

The soldering system disclosed herein is equipped with a temperature measuring device positioned above the jet nozzle, allowing the soldering of circuit boards to be performed with consistent quality.

1 is a schematic diagram showing a soldering system 1 according to a first embodiment. FIG. 2 is a diagram showing the configuration of a control device 9. FIG. 2 is a diagram showing the configuration of a first learning device 22. 1A is a diagram showing an example of temperature distribution data on the surface of a substrate 10. FIG. 1B is a diagram showing an example of target temperature distribution data on the surface of a substrate 10. FIG. 1C is a diagram showing an example of temperature difference distribution data on the surface of a substrate 10. FIG. 13 is a diagram showing a classification image in a support vector machine. 2 is a diagram showing the configuration of a second learning device 23. FIG. FIG. 1 is a diagram illustrating an example of a configuration of a neural network. 2 is a diagram showing the configuration of a first inference device 24. 2 is a diagram showing the configuration of a second inference device 25. FIG. 4 is a flowchart showing a learning procedure performed by a first learning device 22 and a second learning device 23 in the first embodiment. 4 is a flowchart showing an inference procedure by a first inference device 24 and a second inference device 25 in the first embodiment. 10 is a flowchart showing a learning procedure performed by a first learning device 22 and a second learning device 23 in the second embodiment. 13 is a flowchart showing an inference procedure by a first inference device 24 and a second inference device 25 in the second embodiment. 13 is a flowchart showing the procedure of inference and relearning by a first inference device 24 and a second inference device 25 in the third embodiment. FIG. 13 is a diagram showing the arrangement of an infrared camera 7 in embodiment 4. 4 is a diagram showing the change over time in the surface temperature of a substrate 10. FIG. 1A is a diagram showing the state when the substrate 10 comes into contact with the molten solder, and FIG. 1B is a diagram showing the state immediately after the substrate 10 is removed from the molten solder. FIG. 13 is a schematic diagram showing a soldering system 1b according to a sixth embodiment. FIG. 13 is a diagram showing the configuration of a soldering system 1a according to a seventh embodiment. 2 is a diagram showing the hardware configuration of a control device 9. FIG.

An embodiment will be described with reference to the drawings.
Embodiment 1.
FIG. 1 is a schematic diagram showing a soldering system 1 according to a first embodiment.

The soldering system 1 comprises a conveying mechanism 2, a flux applicator 3, a preheater 4, a solder bath 5, a fluxer control unit 18, a jet motor 15, a solder bath heater 14, a temperature measuring device 17, a control device 9, and a soldering inspection device 16.

The flux applicator 3 applies flux 11 to the underside (soldering surface) of the board 10 on which the electronic component 201 to be soldered is mounted. Methods for applying the flux 11 include spraying, foaming, and immersion.

The fluxer control unit 18 controls the amount of flux 11 applied. The amount of flux 11 applied is determined by the flow rate of the flux liquid, the pressure of the compressed air, and the movement speed of the two nozzles. The amount of flux 11 applied varies depending on the degree of clogging of the nozzle in the flux applicator 3, the variation in the operation of the flux applicator 3, and the variation in the exhaust volume of the exhaust fan in the flux applicator 3. If the amount of flux 11 applied is large, the time required for the solvent to evaporate in the preheating process in the preheater 4 becomes long, so the temperature of the substrate 10 does not rise sufficiently. If the amount of flux 11 applied is small, the solvent evaporates quickly in the preheating process in the preheater 4, so the temperature of the substrate 10 becomes too high.

The preheater 4 preheats the board 10. The purpose of preheating is to volatilize the solvent in the flux 11 and to heat the board 10 before soldering. This allows the flux 11 to have an oxide film removing effect, improving the rate of successful soldering. The heating methods of the preheater 4 include infrared rays, far infrared rays, and hot air. In some cases, only the bottom surface (soldering surface) of the board 10 is heated, and in other cases, the top surface (non-soldering surface, component surface) of the board 10 is also heated.

The solder bath 5 stores molten solder.
The jet nozzle 13 jets the molten solder 12 stored in the solder bath 5. This brings the underside of the board 10 into contact with the molten solder 12, and soldering is performed. The jet nozzle 13 includes a primary nozzle 13a and a secondary nozzle 13b. The primary nozzle 13a forms rough waves and supplies the molten solder to every corner of the soldering surface of the board 10. The secondary nozzle 13b forms neat waves so that an appropriate amount of solder adheres to the board 10.

The solder bath heater 14 melts the solder.
The jet motor 15 rotates the impeller to send the molten solder 12 to the jet nozzle 13. This causes the molten solder 12 to jet from the jet nozzle 13.

The control device 9 controls each component of the soldering system 1. For example, the control device 9 controls the temperature of the solder bath heater 14, the rotation speed of the jet motor 15, and the angle of the jet nozzle 13.

The temperature measuring device 17 is disposed above the secondary nozzle 13b constituting the jet nozzle 13. The temperature measuring device 17 measures the surface temperature of the substrate 10. The temperature measuring device 17 is equipped with an infrared camera. The temperature measuring device 17 photographs the substrate 10 with the infrared camera. The temperature measuring device 17 measures the temperature distribution on the surface of the substrate 10 based on the captured infrared image. The temperature measuring device 17 outputs information on the measured temperature distribution on the surface of the substrate 10 to the control device 9. The infrared camera may be equipped with a swivel mechanism that can change the direction of the optical axis of the infrared camera. When a large component is mounted on the substrate 10, if the normal optical axis direction is blocked by the large component and the temperature of the surface of the substrate 10 cannot be measured, the optical axis direction can be changed to avoid being blocked by the large component, thereby measuring the surface temperature of the substrate 10.

The temperature measuring device 17 is disposed above the secondary nozzle 13b, so that the temperature measuring device 17 can measure the temperature of the substrate 10 immediately after the substrate 10 is separated from the jetted molten solder. This stabilizes the quality of the soldering to the substrate 10.

The control device 9 is connected to a soldering inspection device 16 .
The soldering inspection device 16 inspects the quality of the soldering of the board 10. Specifically, the soldering inspection device 16 has an image analysis function. The soldering inspection device 16 analyzes the state of the soldering based on the captured image of the board 10. In detail, the soldering inspection device 16 detects the presence or absence of an image corresponding to "icicle-shaped" solder attached to the terminal of the electronic component 201, and the presence or absence of an image corresponding to a solder bridge. The soldering inspection device 16 also obtains the area to which the solder is attached, and evaluates whether this area is excessive or insufficient with respect to a predetermined area. Furthermore, the soldering inspection device 16 evaluates the gloss of the solder and evaluates whether it satisfies the gloss standard. The soldering inspection device 16 evaluates the above-mentioned multiple items. The soldering inspection device 16 inspects the quality of the soldering based on the evaluation of each item.

The control device 9 determines control parameters for keeping the temperature distribution on the surface of the substrate 10 within an acceptable range. The control device 9 controls each part of the soldering system 1 based on the determined control parameters.

Figure 2 shows the configuration of the control device 9. The control device 9 includes an identification and setting unit 21, a storage device 27, and a drive control unit 26.

The identification and setting unit 21 includes a first learning device 22, a second learning device 23, a first inference device 24, and a second inference device 25.

The memory device 27 includes a target temperature distribution memory unit 76, an allowable temperature difference distribution memory unit 73, a first learning data memory unit 71, a second learning data memory unit 74, a first learned model memory unit 72, and a second learned model memory unit 75.

The target temperature distribution memory unit 76 stores target temperature distribution data for the surface of the substrate 10. It is desirable that the temperature distribution on the surface of the substrate 10 onto which the molten solder is sprayed is uniform. The surface temperature distribution measured from above the substrate 10 is not uniform because the electronic components 201 are mounted on the substrate 10. Therefore, the target temperature distribution memory unit 76 stores in advance target temperature distribution data for the surface of the substrate 10 that takes into account the mounted electronic components 201. The target temperature distribution data for the surface of the substrate 10 is determined in advance by experiment. Specifically, the target temperature distribution data for the surface of the substrate 10 is created by statistically processing a plurality of temperature distribution data that did not result in soldering failure.

The allowable temperature difference distribution memory unit 73 stores allowable temperature difference distribution data. The allowable temperature difference distribution data defines the allowable range of the temperature difference at each position of the temperature difference distribution data on the surface of the substrate.

FIG. 3 is a diagram showing the configuration of the first learning device 22. As shown in FIG.
The first learning device 22 includes a temperature distribution data creation unit 51, a temperature difference distribution data creation unit 52, a feature extraction unit 53, a first learning data creation unit 54, and a first trained model generation unit 55.

The temperature distribution data creation unit 51 acquires information on the temperature distribution of the surface of the substrate 10 from the temperature measurement device 17. The temperature distribution data creation unit 51 divides the surface of the substrate 10 into a two-dimensional matrix having an area of, for example, 1000 x 1000. The temperature distribution data creation unit 51 sets coordinates for the matrix from 1 to 1000 in the vertical direction and 1 to 1000 in the horizontal direction. This process makes it possible to associate the temperature and coordinates at each position on the surface of the substrate 10 and process them. The temperature distribution data creation unit 51 identifies the temperature for each position in the matrix from the acquired temperature distribution information and creates temperature distribution data for the surface of the substrate 10.

The temperature difference distribution data creation unit 52 creates temperature difference distribution data on the surface of the substrate 10 which represents the difference between the temperature distribution data on the surface of the substrate 10 and the target temperature distribution data on the surface of the substrate 10 stored in the target temperature distribution memory unit 76.

Figure 4(a) is a diagram showing an example of temperature distribution data on the surface of the substrate 10. Figure 4(b) is a diagram showing an example of target temperature distribution data on the surface of the substrate 10. Figure 4(c) is a diagram showing an example of temperature difference distribution data on the surface of the substrate 10.

4(a) to (c), the area enclosed by an outer frame represents the surface area of the substrate 10. The lines within the frame represent isotherms. For example, isotherms are shown every 5°C.

The feature extraction unit 53 extracts features of the temperature difference distribution data on the surface of the substrate 10. The features are, for example, data such as the absolute value of the temperature, the density of the isotherm, the shape of the isotherm, the amount of change in the isotherm with respect to the temperature distribution on the surface of the substrate 10 measured previously, or the amount of change in the isotherm with respect to the temperature distribution on the surface of the substrate 10 acquired a set time ago. The set time can be, for example, one hour. The features may be data obtained by subjecting these data to principal component analysis or independent component analysis.

The first learning data creation unit 54 receives the features of the temperature difference distribution data on the surface of the board 10 output from the feature extraction unit 53, and receives the inspection results of the quality of the soldering of the board 10 output from the soldering inspection device 16.

The first learning data creation unit 54 creates first learning data for a plurality of boards 10, the first learning data consisting of a set of features of the temperature difference distribution data on the surface of the board 10 and a label (correct answer) indicating whether the soldering on the board 10 is good or bad, and stores the data in the first learning data memory unit 71.

The first learning data storage unit 71 stores the first learning data.
The first trained model generation unit 55 uses the first training data stored in the first training data storage unit 71 to generate a first trained model that estimates the quality of the soldering of the board 10 from the feature quantities of the temperature difference distribution data on the surface of the board 10. The first trained model generation unit 55 stores the generated first trained model in the first trained model storage unit 72.

The first trained model storage unit 72 stores the first trained model.
The first trained model generation unit 55 generates the first trained model by so-called supervised learning, for example, according to a support vector machine. Here, supervised learning refers to a method of providing the first trained data consisting of a set of input and result (label) data to the first trained model generation unit 55 to learn features in the first trained data and infer a result from the input.

Figure 5 shows a classification image in a support vector machine. For ease of explanation, Figure 5 shows it in two dimensions.

The black circles represent the features of the temperature difference distribution data of the board 10 that is included in the first learning data and is judged to have poor soldering. The triangles represent the features of the temperature difference distribution data of the surface of the board 10 that is included in the first learning data and is judged to have good soldering.

In the learning phase, the support vector machine constructs, based on the first learning data, a discrimination surface that distinguishes between features of the temperature difference distribution data of the surface of the board 10 for which the soldering is determined to be poor and features of the temperature difference distribution data of the surface of the board 10 for which the soldering is determined to be good.

In the inference phase, the support vector machine determines whether the soldering of the substrate 10 to be identified is good or bad by identifying which of the two spaces divided by the identification surface the feature quantities of the temperature difference distribution data of the surface of the substrate 10 to be identified belong to. The stars represent examples of feature quantity vectors of the temperature difference distribution data of the substrate 10 to be identified. In this example, the substrate 10 to be identified is determined to be defective.

FIG. 6 is a diagram showing the configuration of the second learning device 23. As shown in FIG.
The second learning device 23 includes a data acquisition unit 61, a control parameter setting unit 62, a simulation unit 63, a second learning data creation unit 64, and a second learned model generation unit 65.

The data acquisition unit 61 acquires temperature difference distribution data on the surface of the substrate 10 when soldering becomes defective from the first learning device 22.

The control parameter setting unit 62 sets the control parameters. When the soldering of the board 10 becomes defective during the learning phase, the control parameter setting unit 62 corrects the control parameters from the standard values by multiple steps. When the control parameter is the temperature of the solder bath heater 14, the control parameter setting unit 62 increases the temperature of the solder bath heater 14 by 5°C, 10°C, and 15°C from the standard value. When the control parameter is the rotation speed of the jet motor 15, the control parameter setting unit 62 increases the rotation speed of the jet motor 15 by 10%, 20%, and 30% from the standard value. When the control parameter is the angle of the jet nozzle 13, the control parameter setting unit 62 increases the angle of the jet nozzle 13 by 10°, 20°, and 30°.

The simulation unit 63 obtains, by simulation, temperature distribution data on the surface of the substrate 10 after the time set after the control parameters are corrected, based on the corrected control parameters output from the control parameter setting unit 62. The simulation unit 63 outputs "temperature difference distribution data on the surface of the substrate 10 after the time set after the control parameters are corrected" which represents the difference between the "temperature distribution data on the surface of the substrate 10 after the time set after the control parameters are corrected" obtained by simulation and the "target temperature distribution data on the surface of the substrate 10" stored in the target temperature distribution storage unit 76.

The second learning data creation unit 64 creates second learning data including multiple sets of temperature difference distribution data on the surface of the substrate 10 when the control parameter is at a standard value, the correction amount of the control parameter, and temperature difference distribution data (correct answer) on the surface of the substrate 10 after a set time has elapsed after the control parameter is corrected, and stores the second learning data in the second learning data memory unit 74.

The second learning data storage unit 74 stores the second learning data.
The second trained model generation unit 65 generates a second trained model that estimates temperature difference distribution data on the surface of the substrate 10 after a set time has elapsed since the control parameter was corrected, from the temperature difference distribution data on the surface of the substrate 10 when the control parameter is at a standard value and the correction amount of the control parameter, using the second training data stored in the second training data storage unit 74. The second trained model generation unit 65 stores the generated second trained model in the second trained model storage unit 75.

The second trained model storage unit 75 stores the second trained model.
The learning algorithm used by the second trained model generation unit 65 may be a known algorithm such as supervised learning, unsupervised learning, or reinforcement learning. As an example, a case where a neural network is applied will be described.

The second trained model generation unit 65 performs so-called supervised learning, for example, according to a neural network model. Here, supervised learning refers to a method of providing pairs of input and result (label) data to the second trained model generation unit 65, learning the features of the training data, and inferring the result from the input.

FIG. 7 is a diagram showing an example of a configuration of a neural network.
A neural network is composed of an input layer made up of a number of neurons, an intermediate layer (hidden layer) made up of a number of neurons, and an output layer made up of a number of neurons. The intermediate layer may be one layer, or two or more layers.

For example, in a three-layered neural network as shown in Figure 7, when multiple inputs are input to the input layer (X1 to Xn), the values are multiplied by weight W1 and input to the intermediate layer (Y1 to Ym), and the result is further multiplied by weight W2 and output from the output layer (Z1 to Zs). This output result changes depending on the values of weights W1 and W2.

In this embodiment, the neural network learns, in accordance with the first learning data, by so-called supervised learning, to output temperature difference distribution data on the surface of the substrate 10 after a set time has elapsed after the control parameters have been modified.

That is, the neural network inputs into the input layer the temperature difference distribution data on the surface of the substrate 10 when the control parameter is at the standard value and the correction amount of the control parameter, and performs learning by adjusting the weights W1 and W2 so that the result output from the output layer approaches the temperature difference distribution data (correct answer) on the surface of the substrate 10 after a set time has elapsed after the control parameter is corrected.

FIG. 8 shows the configuration of the first inference device 24. As shown in FIG.
The first inference device 24 includes a control parameter setting unit 36 , a temperature distribution data creation unit 31 , a temperature difference distribution data creation unit 32 , a feature extraction unit 33 , a first estimation unit 34 , and a data output unit 35 .

The control parameter setting unit 62 sets the control parameters to standard values.
Similar to the temperature distribution data creation unit 51 in the first learning device 22 , the temperature distribution data creation unit 31 acquires information on the temperature distribution on the surface of the substrate 10 from the temperature measurement device 17 and creates temperature distribution data on the surface of the substrate 10 .

The temperature difference distribution data creation unit 32, similar to the temperature difference distribution data creation unit 52 in the first learning device 22, creates temperature difference distribution data on the surface of the substrate 10 which represents the difference between the temperature distribution data on the surface of the substrate 10 and the target temperature distribution data on the surface of the substrate 10 stored in the target temperature distribution memory unit 76.

The feature extraction unit 33 extracts features of the temperature difference distribution data on the surface of the substrate 10, similar to the feature extraction unit 53 in the first learning device 22.

The first estimation unit 34 reads out a first trained model that estimates the quality of the soldering of the board 10 from the feature quantities of the temperature difference distribution data on the surface of the board 10 from the first trained model storage unit 72. The first estimation unit 34 obtains data representing the quality of the soldering of the board 10 by inputting the feature quantities of the temperature difference distribution data on the surface of the board 10 output from the feature quantity extraction unit 33 into the first trained model.

When the first estimation unit 34 estimates that the soldering of the board 10 is defective, the data output unit 35 outputs temperature difference distribution data on the surface of the board 10 when the control parameters generated by the temperature difference distribution data creation unit 32 are standard values.

FIG. 9 shows the configuration of the second inference device 25. As shown in FIG.
The second inference device 25 includes a data acquisition unit 41 , a second estimation unit 42 , and a control parameter setting unit 43 .

The control parameter setting unit 43 sets the control parameter to a standard value or a value modified from the standard value by the modification amount.

The data acquisition unit 41 acquires temperature difference distribution data on the surface of the substrate 10 when the control parameters output from the data output unit 35 of the first inference device 24 are standard values.

The second estimation unit 42 reads out from the second trained model memory unit 75 a second trained model that estimates the temperature difference distribution data on the surface of the substrate 10 after a set time has elapsed after the control parameter is corrected, from the temperature difference distribution data on the surface of the substrate 10 when the control parameter is at a standard value and the correction amount of the control parameter.

The second estimation unit 42 obtains temperature difference distribution data on the surface of the substrate 10 when the control parameter output from the data acquisition unit 41 is at a standard value and the correction amount of the control parameter corrected by the control parameter setting unit 43 into the second learned model, thereby obtaining temperature difference distribution data on the surface of the substrate 10 after a set time has elapsed after the control parameter is corrected.

Figure 10 is a flowchart showing the learning procedure by the first learning device 22 and the second learning device 23 in embodiment 1.

In step S101, a drive switch (not shown) is set to ON. This causes the conveying mechanism 2 to convey the substrate 10 to the top of the solder bath 5.

In step S102, the control parameter setting unit 62 sets the control parameters to standard values.

In step S103, the temperature distribution data creation unit 51 creates temperature distribution data on the surface of the substrate 10.

In step S104, the temperature difference distribution data creation unit 52 creates temperature difference distribution data on the surface of the substrate 10 which represents the difference between the temperature distribution data on the surface of the substrate 10 and the target temperature distribution data on the surface of the substrate 10 stored in the target temperature distribution memory unit 76.

In step S105, the feature extraction unit 53 extracts features of the temperature difference distribution data on the surface of the substrate 10.

In step S106, the soldering inspection device 16 inspects the quality of the soldering on the board 10.

In step S107, the first learning data creation unit 54 adds a set of the features of the temperature difference distribution data on the surface of the board 10 obtained in step S105 and the label (correct answer) of whether the soldering of the board 10 is good or bad obtained in step S106 to the first learning data in the first learning data memory unit 71.

If an end instruction is input in step S108, the process proceeds to step S109. If an end instruction is not input, the process returns to step S103, and steps S103 to S107 are repeated for the next substrate 10 to be transported.

In step S109, a drive switch (not shown) is set to off, thereby stopping the transport of the substrate 10 by the transport mechanism 2.

In step S110, the first trained model generation unit 55 uses the first training data stored in the first training data storage unit 71 to generate a first trained model that estimates the quality of the soldering of the board 10 from the feature quantities of the temperature difference distribution data on the surface of the board 10. The first trained model generation unit 55 stores the generated first trained model in the first trained model storage unit 72.

In step S111, the data acquisition unit 61 of the second learning device 23 acquires from the first learning device 22 one piece of temperature difference distribution data on the surface of the substrate 10 when soldering becomes defective.

In step S112, the control parameter setting unit 62 modifies the control parameters from the standard values by ΔP, 2×ΔP, 3×ΔP, ... and n×ΔP.

In step S113, the simulation unit 63 obtains temperature distribution data on the surface of the substrate 10 after the time set after the control parameters are corrected by simulation based on the corrected control parameters output from the control parameter setting unit 62. The simulation unit 63 outputs temperature difference distribution data on the surface of the substrate 10 after the time set after the control parameters are corrected, which represents the difference between the temperature distribution data on the surface of the substrate 10 after the time set after the control parameters are corrected and the target temperature distribution data on the surface of the substrate 10 stored in the target temperature distribution storage unit 76. For the corrected n control parameters, temperature difference distribution data on the surface of the substrate 10 after the time set after the control parameters are corrected is output.

In step S114, the second learning data creation unit 64 adds to the second learning data in the second learning data storage unit 74 a set of temperature difference distribution data on the surface of the substrate 10 when the control parameters acquired in step S111 are standard values, the correction amount of the control parameters acquired in step S112, and temperature difference distribution data (correct answer) on the surface of the substrate 10 after a set time has elapsed since the control parameters acquired in step S113 were corrected. The above set is added to the second learning data for the n corrected control parameters.

If an end command is input in step S115, the process proceeds to step S116. If an end command is not input, the process returns to step S111, and steps S111 to S114 are repeated for another board 10 with defective soldering.

In step S116, the second trained model generation unit 65 uses the second training data stored in the second training data storage unit 74 to generate a second trained model that estimates the temperature difference distribution data on the surface of the substrate 10 after a set time has elapsed since the control parameter was corrected, from the temperature difference distribution data on the surface of the substrate 10 when the control parameter is at a standard value and the correction amount of the control parameter. The second trained model generation unit 65 stores the generated second trained model in the second trained model storage unit 75.

Figure 11 is a flowchart showing the inference procedures by the first inference device 24 and the second inference device 25 in embodiment 1.

In step S201, a drive switch (not shown) is set to ON. This causes the conveying mechanism 2 to convey the substrate 10 to the top of the solder bath 5.

In step S202, the control parameter setting unit 36 sets the control parameters to standard values.

In step S203, the temperature distribution data creation unit 31 creates temperature distribution data on the surface of the substrate 10.

In step S204, the temperature difference distribution data creation unit 32 creates temperature difference distribution data on the surface of the substrate 10 which represents the difference between the temperature distribution data on the surface of the substrate 10 and the target temperature distribution data on the surface of the substrate 10 stored in the target temperature distribution memory unit 76.

In step S205, the feature extraction unit 33 extracts features of the temperature difference distribution data on the surface of the substrate 10.

In step S206, the first estimation unit 34 reads out from the first trained model memory unit 72 a first trained model that estimates the quality of the soldering of the substrate 10 from the features of the temperature difference distribution data on the surface of the substrate 10.

In step S207, the first estimation unit 34 obtains data representing the quality of the soldering of the board 10 by inputting features of the temperature difference distribution data on the surface of the board 10 output from the feature extraction unit 33 into the first learned model.

If the soldering of the board 10 is defective in step S208, the process proceeds to step S209. If the soldering of the board 10 is good, the process ends.

In step S209, the data acquisition unit 41 acquires temperature difference distribution data on the surface of the substrate 10 when the control parameters created by the temperature difference distribution data creation unit 32 in step S204 are standard values.

In step S210, the second estimation unit 42 reads out from the second trained model memory unit 75 a second trained model that estimates the temperature difference distribution data on the surface of the substrate 10 after a set time has elapsed after the control parameter is corrected, from the temperature difference distribution data on the surface of the substrate 10 when the control parameter is at a standard value and the correction amount of the control parameter.

In step S211, the second estimation unit 42 sets K=1.
In step S212, the second estimator 42 sets the correction amount of the control parameter to K×ΔP.

In step S213, the second estimation unit 42 obtains temperature difference distribution data of the surface of the substrate 10 when the control parameter output from the data acquisition unit 41 is at a standard value and the correction amount of the control parameter into the second learned model, thereby obtaining temperature difference distribution data of the substrate after a set time has elapsed after the control parameter is corrected.

In step S214, the second estimation unit 42 determines that the temperature difference distribution data on the surface of the substrate 10 is within the allowable range when the temperature difference at each position of the temperature difference distribution data on the surface of the substrate 10 after the lapse of a set time period after the control parameters are corrected is equal to or smaller than the temperature difference at each position of the allowable temperature difference distribution data in the allowable temperature difference distribution storage unit 73. The second estimation unit 42 determines that the temperature difference distribution data on the surface of the substrate 10 is outside the allowable range when the temperature difference at each position of the temperature difference distribution data on the surface of the substrate 10 after the lapse of a set time period after the control parameters are corrected exceeds the temperature difference at each position of the allowable temperature difference distribution data. When the temperature difference distribution data on the surface of the substrate 10 is within the allowable range, the process proceeds to step S216, and when it is outside the allowable range, the process proceeds to step S215.

In step S215, the second estimation unit 42 increments K by 1. Then, the process returns to step S211.

In step S216, the control parameter setting unit 43 modifies the control parameters by the modification amount. The modified control parameters are output to the drive control unit 26. The drive control unit 26 drives the soldering system 1 based on the modified control parameters. This causes, for example, the accumulated dross to melt or the position where the dross is accumulated to move. Then, the dross is uniformly sprayed from the jet nozzle 13, and the temperature distribution on the surface of the board 10 falls within the range of the allowable temperature distribution.

Embodiment 2.
Fig. 12 is a flowchart showing a learning procedure by first learning device 22 and second learning device 23 in embodiment 2. Steps S101 to S111 in the flowchart of Fig. 12 are similar to steps S101 to S111 in the flowchart of Fig. 10, and therefore description thereof will not be repeated.

In step S312, the control parameter setting unit 62 sets K=1.
In step S313, the control parameter setting unit 62 modifies the control parameter by K×ΔP from the standard value.

In step S314, the simulation unit 63 obtains, by simulation, temperature distribution data on the surface of the substrate 10 after the time set after the control parameters are corrected, based on the corrected control parameters output from the control parameter setting unit 62. The simulation unit 63 outputs temperature difference distribution data on the surface of the substrate 10 after the time set after the control parameters are corrected, which represents the difference between the temperature distribution data on the surface of the substrate 10 after the time set after the control parameters are corrected and the target temperature distribution data on the surface of the substrate 10 stored in the target temperature distribution storage unit 76.

In step S315, the second learning data creation unit 64 determines that the temperature difference distribution data on the surface of the substrate 10 is within the allowable range when the temperature difference at each position of the temperature difference distribution data on the surface of the substrate 10 after the lapse of a set time period after the control parameters are corrected is equal to or smaller than the temperature difference at each position of the allowable temperature difference distribution data in the allowable temperature difference distribution storage unit 73. The second learning data creation unit 64 determines that the temperature difference distribution data on the surface of the substrate 10 is outside the allowable range when the temperature difference at each position of the temperature difference distribution data on the surface of the substrate 10 after the lapse of a set time period after the control parameters are corrected exceeds the temperature difference at each position of the allowable temperature difference distribution data. When the temperature difference distribution data on the surface of the substrate 10 is within the allowable range, the process proceeds to step S317, and when it is outside the allowable range, the process proceeds to step S316.

In step S316, the control parameter setting unit 62 increments K by 1. Processing then returns to step S313.

In step S317, the second learning data creation unit 64 adds a set of the temperature difference distribution data on the surface of the substrate 10 when the control parameter acquired in step S111 is at the standard value and the correction amount (correct answer) of the control parameter set in step S313 to the second learning data in the second learning data memory unit 74.

If an end command is input in step S318, the process proceeds to step S319. If an end command is not input, the process returns to step S111, and steps S111 and S312 to S317 are repeated for another board 10 with defective soldering.

In step S319, the second trained model generation unit 65 uses the second training data stored in the second training data storage unit 74 to generate a second trained model that estimates the correction amount of the control parameter from the temperature difference distribution data on the surface of the substrate 10 when the control parameter is at a standard value. The second trained model generation unit 65 stores the generated second trained model in the second trained model storage unit 75.

Figure 13 is a flowchart showing the inference procedure by the first inference device 24 and the second inference device 25 in embodiment 2. Steps S201 to S209 in the flowchart of Figure 13 are similar to steps S101 to S209 in the flowchart of Figure 11, so the description will not be repeated.

In step S410, the second estimation unit 42 reads out from the second trained model memory unit 75 a second trained model that estimates the correction amount of the control parameter from the temperature difference distribution data on the surface of the substrate 10 when the control parameter is at a standard value.

In step S411, the second estimation unit 42 obtains the correction amount of the control parameter by inputting the temperature difference distribution data of the surface of the substrate 10 when the control parameter output from the data acquisition unit 41 is a standard value into the second learned model.

In step S412, the control parameter setting unit 43 modifies the control parameter by the correction amount.

Embodiment 3.
Fig. 14 is a flowchart showing the procedure of inference and relearning by first inference device 24 and second inference device 25 in embodiment 3. Steps S201 to S209 and S410 to S412 in the flowchart of Fig. 14 are similar to steps S201 to S209 and S410 to S412 in the flowchart of Fig. 13, and therefore description thereof will not be repeated.

After step S207, steps S501 and S502 are executed before step S208.

In step S501, the first learning data creation unit 54 adds a set of features of the temperature difference distribution data on the surface of the substrate 10 obtained in step S205 and a label (correct answer) of the quality of the soldering of the substrate obtained in step S207 to the first learning data in the first learning data memory unit 71.

In step S502, the first trained model generation unit 55 uses the first training data stored in the first training data storage unit 71 to update the first trained model that estimates the quality of the soldering of the board 10 from the feature quantities of the temperature difference distribution data on the surface of the board 10. The first trained model generation unit 55 stores the updated first trained model in the first trained model storage unit 72.

Through steps S501 and S502, as operation of the soldering system 1 continues, the amount of accumulated first learning data increases, and as the amount of accumulated first learning data increases, the estimation accuracy of the first trained model increases.

After step S412, steps S503 and S504 are executed.
In step S503, the second learning data creation unit 64 adds a set of the temperature difference distribution data on the surface of the substrate 10 when the control parameter acquired in step S204 is at the standard value and the correction amount (correct answer) of the control parameter acquired in step S411 to the second learning data in the second learning data memory unit 74.

In step S504, the second trained model generation unit 65 uses the second training data stored in the second training data storage unit 74 to update the second trained model that estimates the correction amount of the control parameter from the temperature difference distribution data on the surface of the substrate 10 when the control parameter is at the standard value. The second trained model generation unit 65 stores the updated second trained model in the second trained model storage unit 75.

Through steps S503 and S504, as operation of the soldering system 1 continues, the amount of accumulated second learning data increases, and as the amount of accumulated second learning data increases, the estimation accuracy of the second trained model increases.

Embodiment 4.
In this embodiment, the temperature measuring device 17 includes an infrared camera 7 .

FIG. 15 is a diagram showing the arrangement of infrared cameras 7 in the fourth embodiment.
As shown in FIG. 15, the substrate 10 is transported by the transport mechanism 2 in a direction that differs from the horizontal direction by a certain angle (5±1°).

The infrared camera 7 is disposed so that the direction of the optical axis K of the infrared camera 7 is perpendicular to the surface of the transported substrate 10. In other words, the direction of the optical axis K of the infrared camera 7 is a direction that differs from the horizontal direction by 90°-(5±1)°.

In addition, in the horizontal direction, the infrared camera 7 is installed between a position PA that is 220 mm upstream from the rear end PN of the secondary nozzle 13b in the horizontal direction and a position PB that is 300 mm downstream in the horizontal direction. The reason why the infrared camera 7 is installed in this wide range is that, depending on the structure of the soldering system 1, there are cases where the infrared camera 7 cannot be installed at the rear end PN of the secondary nozzle 13b.

By installing the infrared camera 7 as described above, it is possible to eliminate distortion of the thermal image captured by the infrared camera 7 and perform accurate measurements. Furthermore, by installing the infrared camera 7, it is possible to reduce the area shaded by the components mounted on the board 10, making it easier to secure the area required for temperature measurement.

Embodiment 5.
FIG. 16 is a diagram showing the change in surface temperature of substrate 10 over time.

The solid line represents the temperature change on the surface of the substrate 10 when the substrate 10 is in contact with the molten solder for a long time. At time t1, the substrate 10 separates from the molten solder.

The dashed line indicates the change in temperature of the surface of substrate 10 when the contact time between substrate 10 and the molten solder is short. At time t2, substrate 10 separates from the molten solder .

After the substrate 10 is removed from the molten solder, the temperature difference on the surface of the substrate 10 due to the difference in the contact time between the substrate 10 and the molten solder is greater. By measuring the temperature of the substrate 10 when it is removed from the molten solder, the difference in the amount of heat input to the substrate 10 can be clearly detected.

FIG. 17A is a diagram showing the state when the substrate 10 comes into contact with the molten solder.
FIG. 17B is a diagram showing the state immediately after the substrate 10 is removed from the molten solder.

In this embodiment, the temperature measuring device 17 measures the temperature of the top surface of the substrate 10 after the substrate 10 has separated from the molten solder sprayed from the jet nozzle 13.

Embodiment 6.
FIG. 18 is a schematic diagram showing a soldering system 1b according to the sixth embodiment.

A hole 151 is formed in the exterior 91 of the soldering system 1. The infrared camera 7 included in the temperature measuring device 17 is disposed outside the exterior 91 of the soldering system 1b, and the lens of the infrared camera 7 is disposed in the portion of the hole 151. This makes it possible to prevent the solvent of the volatilized flux from adhering to the infrared camera 7, thereby enabling accurate temperature measurement.

Because the infrared camera 7 is located outside the soldering system 1b, the influence of temperature changes in the infrared camera 7 during maintenance, etc. can be reduced. With a typical infrared camera, if there is a temperature change in the camera body, the measured temperature also changes, but with this configuration, more accurate measurements can be made.

Embodiment 7.
FIG. 19 is a diagram showing the configuration of a soldering system 1a according to the seventh embodiment.

The soldering system 1 a of the seventh embodiment includes a protective window 98 .
The protective window 98 is provided to protect the lens of the infrared camera 7. This makes it possible to prevent flux from adhering to the infrared camera 7 even during long-term operation, so that infrared rays are not blocked and accurate temperature measurement is possible.

Embodiment 8.
FIG. 20 is a diagram showing the hardware configuration of the control device 9.

The control device 9 can be configured with corresponding operations as digital circuit hardware or software. When the functions of the control device 9 are realized using software, the control device 9 includes, for example, a processor 1002 and a memory 1003 connected by a bus 1001, as shown in FIG. 20. The memory 1003 includes a ROM (Read Only Memory) and a RAM (Random Access Memory). The programs executed by the processor 1002 and the data required to execute the programs are stored in the ROM. Data created during program execution is stored in the RAM.

Variant examples.
The present disclosure includes the following modifications.

(1) Position of the infrared camera By installing the infrared camera included in the temperature measuring device 17 away from directly below the exhaust duct (not shown) provided in the soldering system 1, it is possible to prevent flux vapor from adhering to the lens of the infrared camera.

(2) Creation of Temperature Distribution Data The temperature distribution data creation unit divided the surface of the board 10 into a two-dimensional matrix of 1000 x 1000. The division of the surface of the board 10 is not limited to this. The temperature distribution data creation unit may divide the surface of the board 10 depending on the size of the board 10, the analysis accuracy of the temperature distribution, the control accuracy of the soldering system 1, etc. For example, the temperature distribution data creation unit may divide the surface of the board 10 into 50 x 50, 100 x 800, 7000 x 3000, etc.

(3) Soldering Inspection The quality of the soldering may be evaluated by an operator.

(4) Simulation The method of judging the simulation result is not limited to that described in the above embodiment. For example, the temperature distribution obtained by the simulation may be input to a support vector machine to identify whether the temperature distribution belongs to a group of defective soldering or a group of normal soldering.

When the accuracy of the first trained model reaches a certain value or more, re-learning of the first trained model may be stopped. When the accuracy of the second trained model reaches a certain value or more, re-learning of the second trained model may be stopped. By stopping re-learning, the processing load on the control device 9 can be reduced.

(5) Input of the First Trained Model In the above embodiment, the input of the first trained model is the feature of the temperature difference distribution data, but is not limited to this. The input of the first trained model may be data representing the temperature distribution of the substrate. For example, the input of the first trained model may be temperature distribution data created by a temperature distribution data creation unit, or temperature difference distribution data created by a temperature difference distribution data creation unit.

(6) Input of the Second Trained Model In the above embodiment, the input of the second trained model is temperature difference distribution data, but is not limited to this. The input of the second trained model may be any data representing the temperature distribution of the substrate. For example, the input of the second trained model may be temperature distribution data created by a temperature distribution data creation unit.

(7) Temperature Measuring Device The temperature measuring device may use a laser method, an ultrasonic method, or an electromagnetic wave method. The temperature measuring device may include a radiation thermometer instead of an infrared camera.

(8) Learning In the present embodiment, a case where supervised learning is applied to the learning algorithm used by the first trained model generation unit and the second trained model generation unit has been described, but this is not limited to this. As for the learning algorithm, in addition to supervised learning, reinforcement learning, unsupervised learning, semi-supervised learning, or the like can also be applied.

The first trained model generation unit and the second trained model generation unit may generate the first trained model and the second trained model using the first trained data and the second trained data created in a plurality of soldering systems. The first trained model generation unit and the second trained model generation unit may acquire the first trained data and the second trained data from a plurality of soldering systems used in the same area, or may use the first trained data and the second trained data collected from a plurality of soldering systems operating independently in different areas.

It is also possible to add or remove soldering systems that collect the first and second trained models during the process. Furthermore, the first and second trained models generated for a soldering system may be applied to another soldering system, and the first and second trained models may be updated by re-learning.

The learning algorithm used in the first trained model generation unit and the second trained model generation unit may be deep learning, which learns to extract features themselves, or machine learning may be performed according to other known methods, such as genetic programming, functional logic programming, or support vector machines.

The first learning device, the first inference device, the second learning device, and the second inference device may be built into the soldering system. Furthermore, the first learning device, the first inference device, the second learning device, and the second inference device may be present on a cloud server.

The first inference device and the second inference device may obtain a first trained model and a second trained model from an external source, such as another soldering system, and perform inference based on these.

(9) Combination of the Embodiments The above-described embodiments may be combined in any desired manner.

The embodiments disclosed herein should be considered to be illustrative and not restrictive in all respects. The scope of the present disclosure is indicated by the claims, not the above description, and is intended to include all modifications within the meaning and scope of the claims.

1, 1a, 1b soldering system, 2 conveying mechanism, 3 flux applicator, 4 preheater, 5 solder bath, 7 infrared camera, 9 control device, 10 substrate, 11 flux, 13 jet nozzle, 13a primary nozzle, 13b secondary nozzle, 14 solder bath heater, 15 jet motor, 16 soldering inspection device, 17 temperature measurement device, 18 fluxer control device, 21 identification and setting device, 22 first learning device, 23 second learning device, 24 first inference device, 25 second inference device, 26 drive control device, 27 storage device, 31, 51 temperature distribution data creation device, 32, 52 temperature difference distribution data creation device, 33, 53 feature extraction device, 34 first estimation device, 35 data output device, 36, 43, 62 control parameter setting device, 41, 61 Data acquisition unit, 42 second estimation unit, 54 first learning data creation unit, 55 first learned model generation unit, 63 simulation unit, 64 second learning data creation unit, 65 second learned model generation unit, 71 first learning data storage unit, 72 first learned model storage unit, 73 allowable temperature difference distribution storage unit, 74 second learning data storage unit, 75 second learned model storage unit, 76 target temperature distribution storage unit, 91 exterior, 98 protective window, 151 hole, 201 electronic component, 1001 bus, 1002 processor, 1003 memory, K optical axis.

Claims (3)

1. A soldering system comprising:
a solder bath for storing molten solder;
a jet nozzle that jets the molten solder in the solder bath toward a substrate;
a conveying mechanism for conveying the substrates successively above the solder bath;
a temperature measuring device disposed above the jet nozzle for measuring a temperature of the upper surface of the substrate;
a drive control unit that controls the soldering system based on a control parameter;
a first learning data storage unit that stores first learning data including a plurality of sets of data representing a difference between a temperature distribution of the board transported by the transport mechanism based on the temperature measured by the temperature measuring device when the control parameter is a standard value and a target temperature distribution, and data representing the quality of the soldering of the board transported by the transport mechanism, which represents an inspection result by a soldering inspection device;
a first learning device that uses the first learning data to generate a first trained model that estimates the quality of soldering on a board from data representing a difference between a temperature distribution on the board and the target temperature distribution;
a second learning data storage unit that stores second learning data including a plurality of sets of data obtained from the first learning device, the data representing the difference between the temperature distribution of the board transported by the transport mechanism when soldering is defective and a target temperature distribution, and the correction amount of the control parameter that brings the difference between the temperature distribution of the board transported by the transport mechanism and the target temperature distribution, obtained by simulation, into an allowable range;
A soldering system comprising: a second learning device that uses the second learning data to generate a second trained model that estimates a correction amount for the control parameter from only data representing the difference between the temperature distribution of the board when the control parameter is at a standard value and the target temperature distribution. The soldering system of claim 1, wherein the second learning device acquires data representing the difference between the temperature distribution of the board when soldering is defective and the target temperature distribution, corrects the control parameters from their standard values, and based on the corrected control parameters, simulates the difference between the temperature distribution of the board transported by the transport mechanism and the target temperature distribution, and creates the second learning data by gradually changing the amount of correction of the control parameters until the difference between the temperature distribution of the board and the target temperature distribution falls within an acceptable range, thereby determining the amount of correction of the control parameters that will bring the difference between the temperature distribution of the board and the target temperature distribution into an acceptable range. 1. A soldering system comprising:
a solder bath for storing molten solder;
a jet nozzle that jets the molten solder in the solder bath toward a substrate;
a conveying mechanism for conveying the substrates successively above the solder bath;
a temperature measuring device disposed above the jet nozzle for measuring a temperature of the upper surface of the substrate;
a drive control unit that controls the soldering system based on a control parameter;
a first storage unit that stores a first trained model that estimates the quality of soldering of a board from data representing a difference between a temperature distribution of the board and a target temperature distribution;
a first inference device that generates data representing a difference between a temperature distribution of the board transported by the transport mechanism and a target temperature distribution based on the temperature measured by the temperature measuring device when the control parameter is a standard value, estimates the quality of soldering of the board transported by the transport mechanism from the generated data representing the difference between the temperature distribution of the board and the target temperature distribution using the first trained model stored in the first memory unit, and outputs data representing the difference between the temperature distribution of the board when the control parameter is a standard value and the target temperature distribution when it is estimated that the soldering of the board is defective;
a second storage unit that stores a second trained model that estimates data representing a difference between a temperature distribution of the substrate when the control parameter is a standard value and a target temperature distribution, and data representing a difference between a temperature distribution of the substrate after a set time has elapsed after the control parameter is corrected and the target temperature distribution, from a correction amount of the control parameter; and
a second inference device that acquires from the first inference device data representing a difference between the temperature distribution of the substrate when the control parameter is a standard value and the target temperature distribution, generates a correction amount for the control parameter, and estimates data representing a difference between the temperature distribution of the substrate after a set time has elapsed after the control parameter is corrected and the target temperature distribution, from the acquired data representing the difference between the temperature distribution of the substrate and the target temperature distribution and the generated correction amount for the control parameter, using the second trained model stored in the second storage unit;
A soldering system, wherein the temperature difference distribution data on the surface of the board output by the first inference device is temperature difference distribution data on the surface of the board when soldering becomes defective .

Images