ELECTRONICS DESIGN
The evolution of AI from assistant to decision-maker
Dr. Jörg Schambach and Phillip Drechsler both at GOEPEL Electronic explore AI in automated optical and X-ray inspection systems
A
rtificial intelligence (AI) aims to increase the efficiency and effectiveness of industrial processes. Ultimately, this should reduce production costs and production time, improve quality and increase the robustness of industrial processes.
The call for more flexible and autonomous automation will become louder and louder - due to increasing demands on product variety, process flexibility or higher costs. This is especially true for the inspection systems that are standard in modern production lines. It is certain that AI is an important factor in this process of change. But where exactly are the potential uses of AI-based methods in the domain of automatic inspection systems? Among other things, the focus is on manual processes such as the creation of inspection programmes or the classification of defects at the verification station, as the cost aspect quickly comes into play there. With this in mind, the software of the AOI systems from GOEPEL electronic was expanded at an early stage to include an AI-based expert system for automated test programme generation. With the software module “MagicClick”, test programmes are created and optimised fully, automatically. The special feature: without any library entry, a production-ready test programme including component library is generated in just a few minutes. The parameters are also adjusted completely, automatically, even taking into account real process fluctuations.
The transformation of AI from assistant to decision-maker
Both in production lines for electronic SMD assemblies (“Surface Mounted Devices”) and in THT production lines (Through Hole Technology), automatic optical inspection or automated X-ray inspection is part of the standard processes for quality assurance. The corresponding inspection systems test the assemblies 100% for correct placement and soldering of the components and defective assemblies are sorted out. In the case of rejected assemblies, it is typical that the defects detected by the inspection system are finally evaluated and classified by human eyes at a verification station. This visual assessment
22
Figure 1: Potentials for the use of AI-based meth- ods in the domain of automatic inspection systems
is a monotonous activity and carries the risk of errors due to fatigue and human error. The danger of wrong classification decisions is increased when complex defect patterns have to be assessed with changing process parameters. Wrong decisions are fatal when an AOI or AXI system has detected an actual defect and this defect is subsequently classified as a “pass”. In this case, we speak of a false- positive classification, which in turn is equivalent to a slip. The defective PCB would be processed further and in the worst case (if the subsequent electrical tests also do not lead to failure) would be delivered to the customer.
This is where the AI advisor software module, newly developed by GOEPEL electronic and integrated into the PILOT Verify verification software, comes in. For each error detected by the AOI or AXI, the AI-based function forms its own decision. In a first stage (Level 1), an assistance function is provided with this additional information. When the operator has made his decision, there are two independent opinions for each error found, analogous to a four-eye principle - that of the operator and that of the AI. If the AI comes to a different conclusion than the operator, a message is displayed and the user is asked to review his decision again.
The assistance function of the AI advisor DECEMBER/JANUARY 2023 | ELECTRONICS TODAY Figure 2: AI advisor
realised in the first stage (Level 1) ultimately ensures that incorrect decisions are prevented during defect verification and that no detected defects are subsequently classified as “pass”. By constantly adding further relevant training data during operation and the subsequent training processes, the AI becomes better and better at making classification decisions. In level 2, the artificial intelligence is then trained to such an extent that all possible error situations are reliably recognised and verification can take place automatically. The artificial intelligence then makes the basic decision and classifies the occurring errors independently. Verification by the operator is only necessary in exceptional situations, namely when the AI cannot make a safe classification decision. The automated classification of almost all faults in level 2 significantly reduces the workload of the operators at the verification stations.
Good training as a basis The application of Deep Learning (DL) is not limited to the pure development of an AI model. For an industrial application of DL methods, the creation of a balanced and valid training database is essential. This training database
Page 1 |
Page 2 |
Page 3 |
Page 4 |
Page 5 |
Page 6 |
Page 7 |
Page 8 |
Page 9 |
Page 10 |
Page 11 |
Page 12 |
Page 13 |
Page 14 |
Page 15 |
Page 16 |
Page 17 |
Page 18 |
Page 19 |
Page 20 |
Page 21 |
Page 22 |
Page 23 |
Page 24 |
Page 25 |
Page 26 |
Page 27 |
Page 28 |
Page 29 |
Page 30 |
Page 31 |
Page 32 |
Page 33 |
Page 34 |
Page 35 |
Page 36 |
Page 37 |
Page 38 |
Page 39 |
Page 40 |
Page 41 |
Page 42 |
Page 43 |
Page 44
