KR20230117957A - Method for checking the abnormality of reflow devices - Google Patents

Abstract

The present invention relates to a method for checking whether a reflow equipment is abnormal.

Description

Method for checking the abnormality of reflow devices {Method for checking the abnormality of reflow devices}

The present invention relates to the field of IT technology, and more particularly, to a method for checking whether a device is abnormal through a voltage value of a servo motor in a reflow device composed of several steps.

The reflow process connects the printed circuit board (PCB) to the component in order to electrically bond the component to the printed circuit board (PCB). A type of soldering process in which high-temperature heat is applied. In general, cream solder is printed and the cream solder is melted while the printed circuit board (PCB) on which the metal-to-metal contact is mounted passes through a reflow device. This includes the process of soldering.

To this end, the reflow equipment is configured as a horizontal tunnel with open sides, and a conveyor with PCBs passes through it. High-temperature heat is provided in the reflow equipment, and soldering is performed while the PCB placed on the conveyor moves along the conveyor.

The inside of the reflow equipment is divided into several steps (PV1 ~ PVn, where n is a natural number) so that soldering can be performed, and a gripper to hold the PCB is provided at each step, which is operated by a servo motor. In addition, productivity is controlled because a plurality of products (mainly, PCBs) can be soldered while moving along the conveyor at one time. Therefore, a problem may occur in the soldering of the product when the gripper or servomotor of each step is abnormal, and in other words, if it is confirmed that the servomotor voltage value of a specific step is outside the appropriate range (ie, a value outside the envelope) If so, it is possible to predict that the gripper or servomotor of the corresponding stage has already had a problem or that the problem will occur in the near future.

Figure 1a shows the voltage values measured by a total of 28 servomotors while the reflow equipment consisting of a total of 28 steps (PV1 to PV28) is operating hundreds of times. The X-axis is each step, and the Y-axis is the voltage value, and the unit is mA. The more observed voltage values (that is, the higher the number of counts) are shown in a wider gray color. See FIG. 2 for a schematic diagram of the reflow equipment used here. The illustrated exemplary reflow equipment 10 consists of a total of 28 steps (PV1 to PV28), each step is provided with a gripper 20 and a servo motor 30 that drives it, the servo motor 30 includes A sensor (S) for measuring the voltage value is located. Product P, primarily a PCB, is shown moving from left to right along a conveyor. In the example shown in the drawing, a total of four products P are transported at one time. On the other hand, FIG. 1B shows the voltage values measured according to the number of products P, and each product P is divided into four cases, eight cases, and twelve cases. In Figures 1A and 1B, the blue color is the average and the red dotted line shows the envelope. A method for calculating the envelope is known in the art, and thus a detailed description thereof will be omitted.

Looking only at the number of counts in any one step in FIGS. 1A and 1B , it can be seen that the three-sigma rule is not followed. Referring to FIG. 4A , this can be seen in more detail. In FIG. 1B , the case where the number of products P is four, the voltage value at the PV15 stage is the X-axis and the number of counts is the Y-axis. As shown, it is clearly shown that the measured values at step PV15 do not follow the 3 sigma rule, and the same is true for other steps.

Since the voltage value of the servomotor 30 of the reflow equipment 10 does not follow the 3 sigma rule, statistical numerical analysis is difficult. Whether or not the equipment is abnormal can be confirmed or predicted through whether there is a value that exceeds the envelope, but since it does not follow the 3 sigma rule, it is not possible to check the abnormality only with the envelope calculated in a general way. For example, looking at the case of step PV23 in FIG. 1A , the upper limit value of the envelope is lower than 0.16 mA. Even though all 0.16mA ~ 0.20mA are the result values that were normally operated, it is judged to be abnormal in the statistical method because it is out of the envelope. In particular, although the number of counts of 0.16mA is relatively high, it is mistakenly judged to be abnormal at all counts.

If statistical numerical interpretation is difficult, it is impossible to apply even if a learning model is created through training. Therefore, it is difficult to determine whether or not there is an abnormality in the current state, and of course, the model predicting that there will be an abnormality is also inaccurate.

(Patent Document 1) Japanese Patent Registration No. 6506219

(Patent Document 2) Japanese Patent Registration No. 6734318

(Patent Document 3) Japanese Laid-open Patent No. 2020-137327

The present invention has been made to solve the above problems.

In reflow equipment, we would like to propose a method that can predict the failure of the equipment with higher accuracy by using the voltage value of the servo motor. As a general statistical method, since the voltage value of the servo motor of the reflow equipment does not follow the 3 sigma rule, we propose a method that can correct it and predict the abnormality using artificial intelligence.

An embodiment of the present invention for solving the above problems is a1) the training data collection module 110 calculates the voltage value of the servomotor 30 and the product (P) of each step (PVn) from the reflow equipment 10 ) Collecting a plurality of training datasets including the number of; - Here, the voltage value is expressed as a value that increases in units of a predetermined value, and different devices are located at each step (PVn) of the reflow equipment 10 (a2) The gap check module 120 determines the training data Identifying a range of voltage values commonly missing in all of a plurality of training datasets as a gap; (a3) mapping, by the mapping module 130, each voltage value to an integer after deleting the range identified as the gap from the training dataset; (a4) clustering, by the clustering module 140, the mapped training dataset based on the number of products (P), and classifying the mapped training dataset into a plurality of clusters; (a5) determining, by the training module 150, a model and an envelope of each cluster by performing training using the mapped training dataset for each cluster; (b1) collecting, by the data collection module 210, a data set including the voltage value of each step (PVn) from the reflow equipment 10; (b2) deleting, by the preprocessing module 220, the range identified as the gap from the collected dataset in the same manner as the step (a3) and then mapping each voltage value to an integer; and (b3) the anomaly checking module 230 determines the cluster of the mapped dataset, checks the envelope of the corresponding cluster model, and then checks whether or not there is a voltage value outside the envelope in the mapped dataset. checking whether or not; Including, it provides a method.

In one embodiment, in the step (b3), (b31) when the abnormality checking module 230 determines that there is a voltage value outside the envelope in the mapped dataset, the step (PVn) of the corresponding voltage value ), and confirming that there is an abnormality in the device corresponding to the identified step in the reflow equipment 10 may be further included.

In one embodiment, in the step (b1), the checked dataset further includes the number of products P, and in the step (b3), (b32) the abnormality checking module 230, The method may further include determining clusters using the number of products (P) included in the mapped dataset.

In one embodiment, in step (b3), (b32) the anomaly checking module 230 compares the degree of change according to each step of the mapped dataset with the degree of change according to each step of all clusters, , determining a cluster having the most similar degree of change as a cluster of the mapped dataset.

In one embodiment, the servo motors 30 of the reflow equipment 10 have preset unused voltage ranges, and in step (a2), (a21) the gap check module 120 A step of confirming the unused voltage value range preset in the reflow equipment 10 as a gap may be further included.

In one embodiment, after the step (b3), when it is confirmed that there is no abnormality in the step (b4) and the (b3), the training module 150 uses a dataset when it is confirmed that there is no abnormality. Further training of the cluster may further include updating the envelope.

By applying the method according to the present invention, it is possible to build a model that satisfies the 3 sigma rule even when there is a gap in which the voltage value is not measured due to the characteristics of the servomotor in the reflow equipment, and by using this, the abnormality can be more accurately Devices at this stage can be identified or predicted.

If the gap can be confirmed from the manufacturer of the reflow equipment or servo motor, it can be input directly, and even if the gap cannot be confirmed, the gap can be automatically estimated and reflected to increase the accuracy of checking whether there is anomaly.

Since the dataset classified as normal is used for model learning again, the accuracy of the result increases as the method according to the present invention is continuously used.

1A and 1B show voltage value measurement results, averages, and envelopes for each step of the reflow equipment operating normally.
Figure 2 is a schematic diagram showing a system for carrying out the method according to the present invention.
3 is a flowchart for explaining a method according to the present invention.
4A to 4C are views for explaining a method of identifying, deleting, and mapping gaps according to the present invention.
5 is an example of a training dataset used in the method according to the present invention.
6 is a diagram showing measurement values processed again after gaps have been deleted and mapped by the method according to the present invention.
7 shows a model clustered and trained with a method according to the present invention.
8 shows a dataset input to a model trained with the method according to the present invention.

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

Description of the system

Referring to Figure 2, a system according to the present invention is described.

A system for performing the method according to the present invention includes a model generating unit 100 and an equipment checking unit 200, which include a plurality of steps PVn, and each step includes a gripper 20 and driving it. It is interlocked with the reflow equipment 10 where the servo motor 30 is located. The servomotor 30 is provided with a sensor S to measure voltage values, and the voltage values measured for each step PVn become a dataset and are training datasets for generating a model or for checking abnormalities. It can be a real dataset. In FIG. 2, the reflow equipment 10 is shown as having a total of 28 steps, but is not limited to that number. Each step is denoted by the form PVn, where n is an integer greater than 1.

The model generation unit 100 generates a model of a cluster to be described later using a training dataset and checks an envelope. It includes a training data collection module 110 , a gap estimation module 120 , a mapping module 130 , a clustering module 140 and a training module 150 .

The training data collection module 110 receives a plurality of voltage values of the servomotor 30 for each step (PVn) from the sensor S of the reflow equipment 10 and performs a function of checking a training data set.

The gap checking module 120 performs a function of manually or automatically checking gaps in the training dataset.

The mapping module 130 performs a function of mapping to integer values after removing gaps from the training dataset.

The clustering module 140 performs a function of classifying a training dataset that has been mapped according to a predetermined criterion into a plurality of clusters.

The training module 150 performs a function of forming a model by performing training for each of a plurality of divided clusters and checking an envelope.

The equipment checking unit 200 checks whether the reflow equipment 10 has an abnormality by acquiring a data set of equipment to be checked for abnormality and inputting the data set into the model generated by the model generating unit 100 . It includes a data input module 210, a pre-processing module 220, and an abnormality checking module 230.

The data input module 210 receives the voltage value of the servomotor 30 for each step (PVn) from the sensor (S) of the reflow equipment 10 to check whether or not there is an abnormality, and collects a data set.

The pre-processing module 220 removes gaps from the dataset collected by the data input module 210 and performs mapping.

The abnormality checking module 230 manually or automatically checks the clusters of the dataset that have been mapped, checks and substitutes a suitable model among the cluster models generated by the model generator 100, and looks at the result, thereby reflow equipment Check if (10) is abnormal.

DESCRIPTION OF THE CONCEPT OF THE INVENTION

The present invention analyzes and reflects the limits of the voltage value of the servomotor 30 of the reflow equipment 10, corrects it with voltage value data that satisfies the 3 sigma rule, and creates a model that determines whether there is an abnormality based on this. do.

As shown in FIG. 4A , the voltage value of the servomotor 30 of the general reflow equipment 10 is expressed as a value that increases by a certain unit. In the illustrated example, it is confirmed that the increase is in units of 0.01 mA, and this unit may vary depending on the equipment or motor.

The voltage value of the servomotor 30 of the reflow equipment 10 exists in a region where the voltage value is not measured. In the equipment shown in FIG. 4A, voltage values of 0.10 mA, 0.11 mA, 0.12 mA, 0.13 mA, 0.14 mA, and 0.15 mA are not measured. This may be a self-property according to the manufacturer of the servo motor 30 . For safe driving of the servomotor 30, it is set not to be driven in a voltage value within a specific range. Alternatively, it may be its own characteristics according to the manufacturer of the reflow equipment 10 . It is set for safe driving of the gripper 20 of the reflow equipment 10. It is commonly confirmed in the servomotor 30 of most of the reflow equipment 10 that there is a region where the voltage value is not measured. In addition, all servomotors 30 provided in one reflow equipment 10 have a characteristic that this non-measurable region is common.

In the present invention, it was confirmed that the 3 sigma rule is not applied due to the gap, which is an unmeasurable area, but the 3 sigma rule is applied when the gap is deleted. Therefore, the dataset is calibrated using this.

The concept proposed by the present invention is to map each voltage value to a specific integer after removing the gap. This is because of the characteristics that an unmeasurable region exists in the voltage value of the servomotor 30 of the reflow equipment 10 and the characteristic that the voltage value increases by a certain unit, and all the servomotors 30 of one reflow equipment 10 ) due to the property that the range of the gap is the same.

As shown in FIG. 4B, after removing gaps from all measured voltage values, each voltage value is mapped to an integer of 1 or more to correct the dataset. The range from 0.10mA to 0.15mA was identified as a gap and eliminated. At the same time, 0.01mA is mapped to 1, 0.02mA is mapped to 2, and so on, removing the gap and 0.16mA is mapped to 10, 0.17mA is mapped to 11, and 0.18mA is mapped to 12. .

After this process, the dataset shown in FIG. 4a is changed to the graph of FIG. 4c. It can be seen that it conforms to the 3 sigma rule. When processing all datasets as shown in Fig. 1a in this way, the 3 sigma rule was satisfied. A model will be trained using the processed dataset, and the dataset measured in the reflow equipment 10 to be checked for anomaly is also processed in the same way.

description of the method

Using Fig. 3 and referring together to Figs. 5 to 8, the method according to the present invention will be described.

First, the training data collection module 110 collects a plurality of training data sets including the voltage value of the servomotor 30 and the number of products P at each step PVn from the reflow equipment 10 ( S110). An example of the collected training dataset is shown in FIG. 5 . In the illustrated example, it can be seen that a total of 912 measurements are performed. The training dataset further contains the number of products (P). This is used in clustering described later.

Next, the gap checking module 120 checks the gap (S120). This can be done manually or automatically. For example, since an unused voltage value may be preset as an equipment standard for each reflow equipment 10 or servomotor 30 , a user may manually input it to the gap checking module 120 . Alternatively, the gap check module 120 may automatically check a voltage value range in which a training dataset is commonly missing in all of a plurality of training datasets as a gap. In the example shown in FIG. 4B, the range from 0.10 mA to 0.15 mA was confirmed as a gap.

Next, the mapping module 130 maps each voltage value to an integer after deleting the range identified as a gap in the training dataset (S130). In the example shown in FIG. 4B, the range from 0.10 mA to 0.15 mA was identified and removed as a gap, and the voltage values in units of 0.01 mA from 0.01 mA to 0.18 mA were mapped to integers from 1 to 12 as other ranges.

The training dataset preprocessed in this way is rearranged by product (P) and statistically processed, as shown in FIG. 6 . The X-axis is the same for each stage (PV1 ~ PV28), but the Y-axis has been changed to a mapped value. Blue represents the mean and red represents the envelope. It can be seen that most of the values during normal operation of the reflow equipment 10 are included in the envelope, except when the product P is 0, that is, when the PCB is not transferred to the conveyor. Because of hundreds of tests, measurement errors were found in some values, and it is natural that these values lie outside the envelope.

Next, the clustering module 140 clusters the mapped training dataset based on the number of products (P) and divides them into a plurality of clusters (S140). This is because the load of the gripper 20 varies according to the number of products P, and thus the voltage value of the servomotor 30 varies.

Thereafter, the training module 150 determines a model and an envelope of each cluster by performing training using the training dataset mapped to each cluster (S150).

7 shows a model of the segmented clusters. Here, the case where the number of products (P) is 1 to 4 is Cluster 0 (Cluster 0), the case of 5 to 8 is Cluster 1 (Cluster 1), and the case of 9 to 12 is classified as Cluster 2 (Cluster 2). . Of course, other methods of classification are also possible. However, it is preferable not to apply in the case of idle where the product P is zero. Red is the average and dark black is the outer envelope.

Since the model has been created, the data set is checked in the reflow equipment 10 to check for abnormalities, and then the abnormalities are checked using the trained model.

The data collection module 210 collects a data set including the voltage value of each step (PVn) from the reflow equipment 10 (S210). 8 shows an example of a dataset.

Next, the preprocessing module 220 deletes the range identified as a gap from the collected dataset in the same manner as described above, and then maps each voltage value to an integer (S220).

Next, the anomaly checking module 230 determines the cluster of the mapped dataset, checks the envelope of the corresponding cluster model (S230), and then checks whether there is a voltage value outside the envelope in the mapped dataset to determine whether there is anomaly. Check (S140). Two methods can be used for cluster determination. First, as shown in FIG. 8, a dataset may include the number of products P, and clusters are determined using this. In the example of FIG. 8, since there are 9 products (P), the cluster 2 model will be used. Second, by comparing the degree of change according to each stage of the mapped dataset and the degree of change according to each stage of all clusters, a cluster having the most similar degree of change may be determined as a cluster of the mapped dataset. For example, a method of examining changes in voltage values in the dataset of FIG. 8 and determining a cluster having the most similar voltage change among Cluster 0, Cluster 1, and Cluster 2 in FIG. 7 as the cluster of the mapped dataset.

When it is confirmed that there is a voltage value outside the envelope in the mapped dataset by the abnormality checking module 230, the step (PVn) of the voltage value is checked and the device corresponding to the step confirmed by the reflow equipment 10 It can be confirmed that there is something wrong with (the gripper 20 or the servo motor 30) or that it may occur in the future (S252).

If it is confirmed by the abnormality checking module 230 that there is a voltage value outside the envelope in the mapped dataset, it is confirmed that the reflow equipment 10 is normal. In this case, in one embodiment of the present invention, when it is confirmed that there is no abnormality, the training module 150 may further train the cluster using the corresponding dataset to update the envelope to increase accuracy (S251).

10: reflow equipment
20: gripper
30: servo motor
100: model generation unit
110: training data collection module
120: gap check module
130: mapping module
140: clustering module
150: training module
200: equipment confirmation unit
210: data input module
220: preprocessing module
230: abnormality check module
S: sensor

Claims (6)

(a1) The training data collection module 110 collects a plurality of training data sets including the voltage value of the servomotor 30 and the number of products P at each step PVn from the reflow equipment 10 step; - Here, the voltage value is expressed as a value that increases in units of a predetermined value, and different devices are located at each stage (PVn) of the reflow equipment 10
(a2) identifying, by the gap checking module 120, a range of voltage values in which the training dataset is commonly missing in all of a plurality of training datasets as a gap;
(a3) mapping, by the mapping module 130, each voltage value to an integer after deleting the range identified as the gap from the training dataset;
(a4) clustering, by the clustering module 140, the mapped training dataset based on the number of products (P), and classifying the mapped training dataset into a plurality of clusters;
(a5) determining, by the training module 150, a model and an envelope of each cluster by performing training using the mapped training dataset for each cluster;
(b1) collecting, by the data collection module 210, a data set including the voltage value of each step (PVn) from the reflow equipment 10;
(b2) deleting, by the preprocessing module 220, the range identified as the gap from the collected dataset in the same manner as the step (a3) and then mapping each voltage value to an integer; and
(b3) The abnormality checking module 230 determines the cluster of the mapped dataset, checks the envelope of the corresponding cluster model, and checks whether there is a voltage value outside the envelope in the mapped dataset, thereby determining whether there is anomaly. Checking; including,
method.
According to claim 1,
In step (b3),
(b31) When it is confirmed that the abnormality checking module 230 has a voltage value outside the envelope in the mapped dataset, it checks the step (PVn) of the voltage value, and the reflow equipment 10 Further comprising the step of confirming that the device corresponding to the identified step has an abnormality,
method.
According to claim 1,
In step (b1), the identified dataset further includes the number of products P,
In step (b3),
(b32) Further comprising the step of determining, by the abnormality checking module 230, a cluster using the number of products (P) included in the mapped dataset,
method.
According to claim 1,
In step (b3),
(b32) The abnormality checking module 230 compares the degree of change according to each step of the mapped dataset with the degree of change according to each step of all clusters, and selects a cluster having the most similar degree of change as the mapped dataset. Further comprising determining as a cluster of
method.
According to claim 1,
The servomotors 30 of the reflow equipment 10 have preset unused voltage ranges,
In step (a2),
(a21) further comprising, by the gap checking module 120, checking the unused voltage value range preset in the reflow equipment 10 as a gap,
method.
According to claim 1,
After the step (b3),
(b4) When it is confirmed that there is no abnormality in step (b3), the training module 150 further trains the cluster using the dataset when it is confirmed that there is no abnormality and updates the envelope. including,
method.

Images