Anritsu机器学习应用于PCB外观检测.pdf - 第3页

Anritsu T echnical Review No.28 September 2020 Application of M achine Learni ng to Printed Ci rcuit Board Ex ternal Inspecti on (3) Figure 3 Examples of SMT Mounting Faults (1) Poor wetting The solder is not spread even…

Anritsu Technical Review No.28 September 2020 Application of Machine Learning to Printed Circuit Board External Inspection

(2)

parts to the board in a single automated process. Generally,

technology for producing a PC board with surface-mounted

parts is called Surface Mount Technology (SMT). As the PC

boards emerge from the reflow oven, the external inspection

check is performed to make sure there are no problems with

the alignment and soldering of parts on boards. This ex-

ternal inspection check can be performed by either people

using the naked eye or a microscope, or by Automated Op-

tical Inspection (AOI) using optical imaging.

At inspection by AOI, an image of the PC board is cap-

tured from a fixed reference position and the solder condi-

tion of each part is assessed from the image; the presence or

absence of parts, their angles, and the part number (printed

on part), etc., are checked to evaluate whether the PC board

is pass or fail. An error report listing the nature of the faulty

part and faulty board is output for failing boards and these

boards are finally checked visually by people to confirm the

fault evaluation.

Since the AOI camera shot is taken from a fixed reference

position, sometimes it may be difficult to visualize the part

leads and electrodes, making it difficult to accurately assess

the part solder conditions. Consequently, PC boards with

parts that cannot be inspected by AOI must depend on in-

spection by human eye.

Recently, chip mounting densities are increasing using

many parts as small as 1.0 × 0.5 mm and 0.6 × 0.3 mm, re-

quiring even more accurate and faster processing using AOI.

Note 1: Solder paste is a high-viscosity mixture of flux and solder

balls with diameters of 20 to 30 µm. It is called solder cream.

Note 2: A solder mask is a stainless-steel stencil sheet that is over-

laid on the PC board; the required amount of solder paste is

painted accurately at required positions on the PC board

through holes cut in the mask.

3 Typical Solder Errors

When parts are mounted correctly, the part lead or elec-

trode is soldered cleanly to the land on the PC board

Note 3

and the external appearance of the joint is a smoothly

curved meniscus. This is described as a fillet. In addition,

the joint surface is "flowing" and "wets" the connected ele-

ments. Figure 2 shows some examples of correct fillet

shapes for solder joints.

If there are any problems with applied solder paste

amount or position, position of mounted parts, or with the

reflow temperature management, the fillet shape may be

deformed, resulting in a poor solder joint between the part

and board. See reference 5 for details of the causes of vari-

ous types of poor SMT joints. Some typical examples are

listed below and shown in Figure 3.

Inspector

Solder

Paste

Adhesi-

Printer

Reflow Oven

External

Inspector

(AOI)

AOI

Terminal

Parts Mounting

PC Board Progress

Reflow

External Check

Solder Paste

Printing and Check

Parts Mounters several units

Adhesive

PC Board

To Next

Process

Visual Inspection

Figure 1 Example of SMT Line

PC Board

Lead

Fillet

Land

Fillet

Land

Electrode

Lead

Fillet

Land

PC Board

(a) Chip Parts (b) SOP and QFP Parts

Figure 2 Examples of Normal Fillet Shapes

Anritsu Technical Review No.28 September 2020 Application of Machine Learning to Printed Circuit Board External Inspection

(3)

Figure 3 Examples of SMT Mounting Faults

(1) Poor wetting

The solder is not spread evenly in the correct amount over

the part leads and electrodes and the fillet form does not

appear wetted. The causes are insufficient solder, leads or

electrodes not in contact with land, and insufficient heat at

reflow.

(2) Solder bridging (shorts)

This problem tends to occur when using very small SOP

and QFP ICs with sizes of 0.5 mm and 0.4 mm

Note 4

or when

the correct amount of solder is not applied to adjacent leads.

The causes are poor solder printing, bent solder leads, and

poor parts mounting.

(3) Vertical chip (gravestone and Manhattan)

This problem occurs when both electrodes of a part are

not soldered simultaneously and surface tension at the end

wetted first causes the chip to stand vertically on one sol-

dered electrode.

The countermeasures are improving the land dimensions

and mounting accuracy, and preheating to reduce the solder

melt time difference.

(4) Non-contact, missing solder

This is caused by the solder paste not melting during re-

flow and remaining in a paste state. The causes are old

solder paste or poor reflow oven temperature control.

Missing solder is caused by lack of solder at the part and

is caused by poor solder paste printing.

(5) Rotated and slipped parts

Rotated or slipped parts are the result of part leads or

electrodes projecting outside the land or from poor posi-

tioning by the parts mounter. Vibration at mounting other

parts or at conveying/transport between processes can re-

sult in surface tension issues causing displacement as in

vertical chip faults.

Note 3: The land is a part where the copper forming the PC board

traces is exposed for soldering to the part leads and elec-

trodes. Sometimes the land surface is gold-plated.

Note 4: SOP and QFP describe IC packages. A SOP (Small Outline

Package) has L-shaped legs from two sides of the rectan-

gular package connecting to lands. A QFP (Quad Flat

Package) has multiple leads on all four sides of the package

connecting to lands.

4 AOI PC Board External Inspection Issues

AOI inspection sets strict evaluation criteria using mul-

tiple parameters so as to not allow fail products to pass in-

spection. As a result, so-called "excess watching" is a com-

mon problem. Excess watching is a phenomenon where

products passing at the visual-inspection level are evalu-

ated as fail by the AOI inspection system. If there is too

much excess watching, visual confirmation after automatic

inspection is increased.

To proactively suppress excess watching, the soldered

part digital evaluation criteria are readjusted repeatedly

over but there can be a problem where excess watching does

not decrease because fine-adjustment cannot be completed

due to momentary changes in the processing conditions

caused by PC board condition and parts mounting ran-

domness.

5 PC Board External Inspection using Machine

Learning

Applying machine learning to the visual confirmation

work following AOI inspection has helped with excess

watching inspection efficiency.

5.1 Leaned Image Data Acquisition

Collecting and annotating (labeling) images used for

machine learning is a key process in applying machine

learning. Consequently, we configured a system (Figure 4)

to capture AOI inspection image data for use as learning

images.

Generally, dedicated terminals are required to capture

AOI inspection results and inspection image data. In this

development, we obtained the terminal interface specifica-

tions from the AOI maker and developed software to di-

rectly capture inspection results and learning images (in-

spection images) at a connected personal computer (PC).

(a) Poor wetting

(b) Solder Bridge

(d) Missing Solder

(e) Rotated Part

Anritsu Technical Review No.28 September 2020 Application of Machine Learning to Printed Circuit Board External Inspection

(4)

This software was run on one Image Training PC on the

AOI LAN (Figure 4).

AOI-related PCs are connected over a dedicated AOI local

area network (LAN) from the viewpoint of higher security

and to prevent non-factory network problems affecting

plant mass-production.

To assure the independence of the AOI LAN, the ma-

chine-learning environment ( and  in Figure 4) was con-

figured on the in-company LAN via a separate independent

LAN (c. Independent LAN in Figure 4) and multiport Net-

work Attached Storage (NAS) ( in Figure 4).

Required data is shared via files on the NAS and the PC

for capturing learning images is programed to send images

and evaluation data stored on the AOI-DB server periodi-

cally to the machine-learning PC.

Generally, at machine learning, only images of pass

products are required, but not images of faulty products.

However, due to disk-space limitations at the AOI-DB serv-

er, there were problems with inability to save images for

products evaluated as pass. Consequently, the focus of at-

tention was only on trouble locations such as electrodes and

leads for images of parts evaluated as fail and machine

learning was used to capture images of fillets of electrodes

and leads of pass parts from within images of fail parts.

When introducing machine evaluation to external inspec-

tion, the learning model obtained from the Ma-

chine-Learning PC was transferred to the Machine Evalua-

tion PC and the result for the evaluated image was saved in

the AOI-DB server as the visual evaluation result that the

Image Training PC input to the AOI terminal. If images

evaluated by AIO as fail were evaluated by machine learn-

ing as pass, the operator in charge of visual evaluation

handled the image as if it had already been visually evalu-

ated.

Reduction of visual inspection and confirmation by eye is

expected with introduction of machine learning because

only parts evaluated as fail by both AOI and machine

learning are inspected.

5.2 Machine Learning Discriminator

Machine learning is a method for allowing a computer to

learn human-like recognition and evaluation. It is per-

formed using the following two processes: Learning, and

Evaluation

1), 2), 3)

(1) Learning Process

Even the best machine-learning algorithm is worthless

without a learning process which must be performed first

using learning data. Learning is repeated over until the

accuracy (learning mistakes) reach the required accuracy

and the learning model is completed.

(2) Evaluation Process

The discriminator (model with completed learning) eval-

uates whether input image data are pass or fail. For exam-

ple, a discriminator that has learned labelled (named)

product images can accurately evaluate the names of prod-

ucts for unknown images that have not been learned.

a. AOI LAN

b. Company LAN

(1) Training image and AOI evaluation result

(2)

Evaluation m

achine learning model

(3) Board evaluation image and result (at machine evaluation introduction)

(1)

(2)

(3)

(1)

c. Inde

endent LAN

 Image

Training PC

(Windows10)



AOI-DB

Server

AOI

Main

AOI

Terminal

AOI

Terminal

AOI

Terminal

 Machine

Learning PC

(Linux)

 Machine

Evaluation

PC (Linux)

Figure

Image Acquisition System Configuration