TWI732296B - System for generating detection model according to standard data to confirm soldering state and method thereof - Google Patents

Abstract

A system for generating a detection model according to standard data to confirm a soldering state and a method thereof are provided. By generating joint classifications in accordance with standard data of solder joints, generating a detection model based on first data corresponding to the solder joints included in every joint classifications, and using the detection model to analyze is any poor soldering in second data detected from a printed circuit board, the system and the method can achieve the effect of reducing number of solder joints misjudged as bridging and shortening required time of reconsideration.

Description

System and method for establishing inspection model based on standard value to confirm welding state

A welding inspection system and method, in particular, refers to a system and method that establishes an inspection model based on standard values to confirm the welding state.

At present, the most commonly used technology in the electronic assembly industry is Surface-Mount Technology (SMT), which means that electronic components such as resistors, capacitors, transistors, and integrated circuits are soldered to a printed circuit board (Printed Circuit Board). , PCB) to form electrical connections. By using surface mount technology, the overall processing speed of the assembly can be increased.

However, due to the miniaturization and increased density of electronic components, the possibility of poor soldering of electronic components on printed circuit boards has increased accordingly. Therefore, the detection of soldering conditions during the manufacturing process of printed circuit boards with any surface mount technology Has become a necessary part.

At present, Solder Paste Inspection (SPI) equipment can detect the volume, area, height, X offset and Y offset of each solder joint, and use the detected index data to determine the solder joint Whether there is poor welding. Therefore, solder paste thickness testing equipment has always been playing a role in inspecting soldering quality, and defects in soldering quality can be found.

But in fact, the accuracy of the solder paste thickness test equipment is not high enough, and it is often possible that half of the solder joints that are judged to be poorly soldered are confirmed by the reviewer that there is no poor soldering. This increases the unnecessary workload of reviewers.

In summary, it can be seen that the prior art has long been a problem of insufficient detection accuracy of the current solder paste thickness testing equipment. Therefore, it is necessary to propose improved technical means to solve this problem.

In view of the problem of insufficient detection accuracy of solder paste thickness testing equipment in the prior art, the present invention discloses a system and method for establishing a detection model based on standard values to confirm the soldering state, in which:

The system for establishing a detection model based on standard values to confirm welding status disclosed in the present invention at least includes: a data collection module for collecting standard value data, first solder joint data, and second solder joint data. The first solder joint The data and standard value data correspond to the solder joints; the solder joint classification module is used to generate solder joint classifications based on the standard value data; the model building module is used to establish inspections based on the solder joint data corresponding to the solder joints included in the solder joint classification Model: The data analysis module is used to use the detection model to analyze whether the second solder joint data has welding defects.

The method for establishing a detection model based on standard values to confirm welding status disclosed in the present invention includes at least the steps of: collecting standard value data and first solder joint data, where both the first solder joint data and standard value data correspond to the solder joints; The standard value data generates the solder joint classification; the detection model is established according to the first solder joint data corresponding to the solder joint included in the solder joint classification; the second solder joint data is collected; the detection model is used to analyze whether the second solder joint data has poor welding.

The system and method disclosed in the present invention are as above. The difference with the prior art is that the present invention generates solder joint classification based on standard value data, and establishes it based on the first solder joint data corresponding to the solder joints included in the solder joint classification After inspecting the model, use the inspection model to analyze whether the second solder joint data generated by the printed circuit board inspection has soldering defects, so as to solve the problems of the previous technology, and can reduce the number of solder joints for manual re-judgment and shorten the manual re-judgment The technical efficacy of the required time.

In the following, the features and implementation of the present invention will be described in detail with the drawings and embodiments. The content is sufficient to enable any person familiar with the relevant art to easily and fully understand the technical means used by the present invention to solve the technical problems and implement them accordingly. The achievable effect of the present invention.

The present invention can generate a detection model based on the standard value data of solder joints or pads (land/pad), and use the detection model to analyze whether the solder joint data has poor soldering. Among them, the standard value data can be the upper and lower limits of the index data of the corresponding solder joint, or the width, height, and area of the corresponding pad. The above index data includes the volume, area, height, horizontal offset, and vertical offset of the corresponding solder joints.

Hereinafter, the system architecture diagram of establishing a detection model based on standard values to confirm the welding state mentioned in the present invention is used to illustrate the operation of the system of the present invention. As shown in "Figure 1", the system of the present invention includes a data collection module 110, a solder joint classification module 120, a model creation module 130, a data analysis module 150, and an output module 160. Among them, the system of the present invention can be applied to the computing device 100.

The data collection module 110 is responsible for collecting data of the first solder joint. The first solder joint data collected by the data collection module 110 is index data corresponding to solder joints on a printed circuit board (PCB). Among them, the printed circuit board usually includes one or more pads, each pad has one or more solder joints, and each solder joint corresponds to a different first solder joint data.

Generally speaking, the data collection module 110 can read the inspection files of the printed circuit board inspected in the past from the storage medium of the computing device 100, and parse the read inspection files to obtain the first recorded in the inspection file. Solder joint data; the data collection module 110 can also be connected to a storage device external to the computing device 100 to download the inspection file of the printed circuit board and obtain the first solder joint data from the inspection file.

The data collection module 110 is also responsible for collecting standard value data. Each standard value data collected by the data collection module 110 may correspond to a different solder joint on the printed circuit board, or it may correspond to one or more solder joints on a different pad included in the printed circuit board .

The data collection module 110 can also obtain the test file and the setting file of the same printed circuit board from the storage medium of the computing device 100 or from a storage device external to the computing device 100, and analyze the obtained setting file, thereby using the setting file Obtain standard value data in.

The data collection module 110 is also responsible for collecting the second solder joint data. Similar to the above, the data collection module 110 can obtain the second solder joint data from the storage medium of the computing device 100 or a storage device external to the computing device 100. The second solder joint data collected by the data collection module 110 includes not only index data corresponding to the corresponding solder joint, but also an image within a certain range around the solder joint (referred to as the inspection image in the present invention). ), can also include location information such as the coordinates or identification data of the solder joint on the pad or printed circuit board.

The second solder joint data collected by the data collection module 110 usually includes and uses solder paste inspection (Solder Paste Inspection, SPI) equipment (not shown in the figure) to detect all solder joints on the printed circuit board. The index data corresponding to the solder joints judged by the testing equipment as poor soldering. However, the present invention is not limited to this, and the second solder joint data may also be index data corresponding to each solder joint on the printed circuit board.

The solder joint classification module 120 is responsible for generating one or more solder joint classifications based on the standard value data collected by the data collection module 110, and each solder joint classification generated by the solder joint classification module 120 includes one or more solder joints. In most embodiments, different solder joint classifications will not include the same solder joint, that is, a solder joint only includes one solder joint classification.

The solder joint classification module 120 can use a specific algorithm to classify solder joints corresponding to the standard value data based on the upper and lower limits of the index data contained in the standard value data. Among them, the algorithm used by the solder joint classification module 120 includes but is not limited to K-means clustering (k-means clustering) and the like.

In some embodiments, the solder joint classification module 120 can adjust the solder joint classification according to the ratio of poor soldering in the second solder joint data analyzed by the data analysis module 150.

The model building module 130 is responsible for creating a detection model based on the solder joint data corresponding to each solder joint included in each solder joint classification generated by the solder joint classification module 120. For example, the model building module 130 can randomly generate k decision trees from the solder joint data, and can use Gini coefficients to select features. Each decision tree continuously traverses all possible feature subsets of the decision tree. In order to find the feature segmentation point with the smallest Gini coefficient to divide the data set into two subsets, until the stop condition is met, the simple voting method is used to use the category with the most votes as the final detection model. However, the manner in which the model building module 130 builds the detection model is not limited to the above.

The model building module 130 can divide the solder joint data into two parts. The model building module 130 can use one part of the solder joint data to train the inspection model, and can use the other part of the solder joint data to predict the accuracy. , Until the accuracy reaches a predetermined value. For example, the model building module 130 can randomly select a predetermined ratio of solder joint data from all solder joint data to train the inspection model, and use the unselected solder joint data to determine the accuracy of the inspection model. When the predetermined value is reached, the model building module 130 can adjust the predetermined ratio, and again randomly select the adjusted solder joint data of the predetermined ratio from all the solder joint data to train the inspection model, and use the unselected solder joint data to determine Check the accuracy of the model. However, the present invention is not limited to this.

The data analysis module 150 is responsible for using the detection model created by the model creation module 130 to analyze whether the second solder joint data collected by the data collection module 110 has welding defects. In more detail, the data analysis module 150 can provide the second solder joint data to the inspection model, so that the detection model generates analysis results corresponding to each second solder joint data. The data analysis module 150 can be based on the data generated by the inspection model. The analysis result determines whether the corresponding second solder joint data indicates whether the corresponding solder joint has poor welding.

The output module 160 is responsible for outputting the second solder joint data with poor welding when the data analysis module 150 determines that the second solder joint data has poor welding. The output module 160 can display the second solder joint data with poor soldering, so that the user can determine the solder joint corresponding to the second solder joint data based on the inspection image contained in the second solder joint data displayed by the output module 160 Whether the welding is indeed bad.

In some embodiments, the output module 160 can also output the position information of the solder joints corresponding to the second solder joint data with poor soldering, for example, the coordinates of the solder joints on the printed circuit board. However, the present invention does not Limited by this.

Next, an embodiment is used to explain the operating system and method of the present invention, and please refer to "Figure 2A" for the flow chart of the method of establishing a detection model based on standard values to confirm the welding state according to the present invention.

First, the data collection module 110 can collect the first solder joint data and the standard value data corresponding to the solder joints on the printed circuit board (step 210). In this embodiment, it is assumed that the data collection module 110 can be connected to a server (not shown in the figure) to download the inspection files and setting files of the printed circuit boards that have been inspected in the past, and analyze the inspection files and setting files respectively, by which Read the first solder joint data and standard value data from the inspection file and the setting file.

After the data collection module 110 collects the first solder joint data and the standard value data (step 210), the solder joint classification module 120 can generate a solder joint classification based on the standard value data collected by the data collection module 110 (step 220). ), the model creation module 130 can create a detection model based on the solder joint data corresponding to the solder joints included in the solder joint classification generated by the solder joint classification module 120 (step 230). In this embodiment, it is assumed that the model building module 130 divides the solder joint data in each solder joint classification into two parts at a ratio of 3:1. Among them, the model building module 130 uses a larger number of solder joints. The point data is used to train the detection model, and a small number of solder joint data is used to evaluate the accuracy of the prediction until the accuracy reaches a predetermined value.

After the model establishment module 130 establishes the inspection model (step 230), the data collection module 110 may collect the second solder joint data (step 240). In this embodiment, it is assumed that the data collection module 110 is connected to the server to download the test report generated by the solder paste thickness test equipment detecting the same printed circuit board, and after analyzing the downloaded test report, it is read from the test report The solder joint data judged by the solder paste thickness test equipment to have poor soldering.

In practice, the model establishment module 130 establishes the inspection model (step 230) and the solder paste thickness test equipment inspects the printed circuit board to generate the inspection report. There is no sequence relationship. That is to say, in the present invention, the model establishment module 130 The detection model can be established before, after or at the same time when the solder paste thickness test equipment generates the test report, and the present invention is not particularly limited.

After the data collection module 110 collects the second solder joint data (step 240), the data analysis module 150 can use the inspection model created by the model creation module 130 to analyze the second solder joint collected by the data collection module 110 Data, the inspection model is used to analyze whether the second solder joint data has poor soldering (step 250). If not, the data analysis module 150 can continue to analyze the next second solder joint data; and if the data analysis module 150 determines that the analysis result output by the inspection model indicates that the second solder joint data has poor soldering, the output module 160 The second solder joint data can be added to the analysis report.

After the data analysis module 150 uses the detection model to analyze whether all the second solder joint data has poor soldering (step 250), the output module 160 can store the generated analysis report in the computing device 100 or the server for output The second solder joint data with poor soldering (step 260) allows the reviewer to re-judge the solder joint corresponding to the second solder joint data based on the analysis report generated by the output module 160. In this embodiment, the output module 160 can also directly display the detection image contained in the second solder joint data to output the second solder joint data with poor soldering (step 260), so as to provide the user with a basis for the displayed detection The image re-judges whether the solder joint corresponding to the second solder joint data is really bad.

In this way, through the present invention, the probability that a solder joint is misjudged as a defective solder can be reduced, and the workload of the re-judgment personnel can be reduced.

In addition, in the above-mentioned embodiment, after the data analysis module 150 uses the model establishment module 130 to establish whether the second solder joint data collected by the data collection module 110 is checked for defective soldering (step 250), it is updated As shown in the process of “Figure 2B”, the model building module 130 may adjust the solder joint classification according to the ratio of the data analysis module 150 to determine that the second solder joint data has poor soldering (step 270).

In summary, it can be seen that the difference between the present invention and the prior art is that after the solder joint classification is generated based on the standard value data, and the inspection model is established based on the first solder joint data corresponding to the solder joint included in the solder joint classification, The detection model is used to analyze whether the second solder joint data generated by the printed circuit board inspection has a technical means for poor soldering. This technical means can solve the problem of insufficient detection accuracy of the solder paste thickness test equipment in the prior art, and then Achieve the technical effect of reducing the number of solder joints for manual re-judgment and shorten the time required for manual re-judgment.

Furthermore, the method of the present invention for establishing a detection model based on standard values to confirm the welding status can be implemented in hardware, software, or a combination of hardware and software, and can also be implemented in a centralized manner in a computer system or distributed with different components Realize in a distributed way of several interconnected computer systems.

Although the embodiments of the present invention are disclosed as above, the content described is not intended to directly limit the scope of patent protection of the present invention. Any person with ordinary knowledge in the technical field to which the present invention belongs, without departing from the spirit and scope of the present invention, makes slight modifications to the form and details of the implementation of the present invention, all belong to the patent protection of the present invention. range. The scope of patent protection of the present invention shall still be determined by the scope of the attached patent application.

100 Computing equipment 110 Data collection module 120 Solder spot classification module 130 Model building module 150 Data analysis module 160 Output module Step 210 Collect standard value data and first solder joint data Step 220 Generate solder joint classification based on standard value data Step 230 Establish an inspection model based on the solder joint data corresponding to the solder joints included in the solder joint classification Step 240 Collect the second solder joint data Step 250 Use the detection model to analyze whether the second solder joint data has poor soldering Step 260 Output the data of the second solder joint with poor soldering Step 270 Adjust the solder joint classification based on the rate of poor soldering in the second solder joint data

Figure 1 is a system architecture diagram of the present invention for establishing a detection model based on standard values to confirm the welding state. Figure 2A is a flow chart of the method of establishing a detection model based on standard values to confirm the welding state according to the present invention. Figure 2B is a flow chart of the method for adjusting the classification of solder joints according to the ratio of poor soldering according to the present invention.

Step 210 Collect standard value data and first solder joint data Step 220 Generate solder joint classification based on standard value data Step 230 Establish an inspection model based on the solder joint data corresponding to the solder joints included in the solder joint classification Step 240 Collect the second solder joint data Step 250 Use the detection model to analyze whether the second solder joint data has poor soldering Step 260 Output the data of the second solder joint with poor soldering

Claims (8)

A method for establishing an inspection model based on standard values to confirm the welding state. The method at least includes the following steps: collecting multiple standard value data and multiple first solder joint data, each of the first solder joint data corresponds to a solder joint, and each A standard value data corresponds to at least one solder joint; at least one solder joint classification is generated based on the standard value data, and each solder joint classification includes at least one solder joint; each solder joint included in each solder joint classification corresponds to Establish a detection model for the first solder joint data; collect at least one second solder joint data; use the detection model to analyze whether the at least one second solder joint data has poor welding; and based on the second solder joint data, there is a poor welding The ratio is adjusted to the classification of the at least one solder joint, and the detection model is re-established to analyze whether the at least one second solder joint has poor welding. For example, the method of establishing an inspection model based on standard values to confirm the welding status as described in item 1 of the scope of patent application, wherein the step of collecting the standard value data is to collect the volume, area, height, horizontal offset, and vertical of each solder joint Offset the upper and lower limits, or collect the width, height, and area of a plurality of pads including at least one solder joint. As described in item 1 of the scope of patent application, the method of establishing a detection model based on standard values to confirm the welding status, in which the step of establishing the detection model based on the solder joint data corresponding to each solder joint included in each solder joint classification further includes using Part of the solder joint data in the solder joint data trains the detection model and uses another part of the solder joint data to predict the accuracy until the accuracy reaches a predetermined value. As described in item 1 of the scope of patent application, the method of establishing an inspection model based on standard values to confirm the soldering status, wherein the step of collecting the at least one second solder joint data is collecting and soldering paste thickness test (Solder Paste Inspection, SPI) equipment The solder joint data corresponding to the solder joints with poor soldering detected. A system for establishing a detection model based on standard values to confirm welding status. The system at least includes: a data collection module for collecting multiple standard value data and multiple first solder joint data, each of the first solder joint data Corresponding to a solder joint, each standard value data corresponds to at least one solder joint, and is used to collect at least one second solder joint data; a solder joint classification module is used to generate at least one solder joint classification based on the standard value data Each of the solder joint classifications includes at least one solder joint; a model creation module for establishing a detection model based on the solder joint data corresponding to each solder joint included in each solder joint classification; and a data analysis module , For using the detection model to analyze whether the at least one second solder joint data has defective welding; wherein, the solder joint classification module is further used to adjust the at least one solder joint according to the ratio of the second solder joint data with defective soldering Point classification, the model building module is further used to rebuild the detection model, so that the data analysis module analyzes again whether the at least one second solder joint has poor soldering. As described in item 5 of the scope of patent application, a system for establishing a detection model to confirm the welding status based on standard values, where the standard value data is the volume, area, height, horizontal offset, and vertical offset of each solder joint The lower limit, or the width, height, and area of multiple pads including at least one of the solder joints. For example, the system for establishing a detection model based on standard values to confirm the welding status as described in item 5 of the scope of patent application, wherein the model establishment module uses part of the solder joint data training Practice the inspection model and use another part of the solder joint data to predict the accuracy until the accuracy reaches a predetermined value. As described in item 5 of the scope of patent application, a system for establishing a detection model to confirm the welding state based on standard values, wherein the at least one second solder joint data is the solder joint corresponding to the solder joint detected by the solder paste thickness test equipment for poor soldering data.

Images