An All-Virtual Approach to Evaluate and Correct Perceived Surface Defects

A discussion with Ing. Arthur Camanho, Smart Manufacturing Director, ESI Group

by Céline Gallerne
image gtc 2019 demo 001 1

Manufacturing a perfect automotive part is not an easy task. But automakers have understood how to identify what counts and what doesn’t, what the human eye can see and what it can’t. Beyond measured quality, the concept of perceived quality is driving efforts (and budgets) for automotive developments. I asked Arthur Camanho, Smart Manufacturing Director, ESI Group, to tell us how leading carmakers are implementing the right technologies to develop an all-virtual, cost-effective, no-frills approach to making auto parts, focusing on perceived quality.

acaArthur Camanho brings over 20 years of experience in mechanical engineering, manufacturing processes, and technical support. He is considered an expert in metal forming simulation and FEA engineering with a concentration on stamping, casting, and welding manufacturing processes.
Camanho earned his mechanical engineering degree from FEI University in Sao Paulo, Brazil. He spent 19 years in the Sao Paulo area working in various capacities for ESI before moving to the North American headquarters to take the lead in the virtual manufacturing arm of the company. Arthur, his wife, and two children reside in Metro Detroit.

Why are parts makers/ automakers taking a digital turn for surface quality control?

In the past, automotive manufacturers would leave quality control for the end of prototyping or even final production. Skilled operators would assess each part sensorially – with eyes and hands – in a “Light Room” under a fluorescent light array to make sure there was no noticeable surface defect. Finding out about defects right at the end of the process meant going all the way back to manufacturing process design, tooling design, or part design as many times as necessary, creating long and costly iteration loops. It wasn’t an effective way to work.

Today, digital solutions are offering automakers a way to skip these tests and prototypes, achieving leaner processes, reducing product development cost and lead times, whilst improving final part quality. There’s a lot at stake when digitalizing surface defects prediction! To get it right, automakers and their technology partners must consider the entire production workflow, including assembly. They must also be able to numerically assess defects as perceived, to tackle the most meaningful surface imperfections.

At what stage of product development or production should OEMs and parts makers introduce quality investigation software?

Numerical solutions deliver quality control prediction on the shop floor, but to maximize the benefits of such technology, it is best to predict perceived defects right from the simulation stage. ESI offers a way for process design engineers to predict surface defects using sheet metal forming simulation, making the assessment directly “as perceived” in a Virtual LightRoom. Of course, the software also predicts all sorts of forming defects such as splits and wrinkles.

Eliminate surface defects by digitalizing every step from simulation to visualizing defects and classifying them by severity levels.

Does this change the way engineers correct visible defects?

The idea is to correct defects earlier, to reduce iteration loops and the time and cost associated. In the best cases, automakers do this with zero real prototypes of a part, or the tooling used to produce it. Those OEMs are making considerable savings.

If quality engineers can decide which defect is most meaningful from the simulation stage, they then go back to the manufacturing process design and do something about it. What is also very important is to spend time-solving defects that will be visible after painting. Sometimes engineers rely on more standard simulation contours to detect surface defects, and those defects will sometimes not show in the real stamped part.

How does software tell the difference between nominal surface quality and perceived surface quality?

At ESI, we have developed an exclusive partnership with Autaza. This Brazilian company co-created with General Motors Brazil a software helping automakers control the quality of stamped automotive parts “as perceived” by the human eye. Autaza Surface utilizes computer vision trained with Artificial Intelligence to predict appearance defects (dents, undulations, damages) before the final design of the part and before the production of the stamping tool. And they are growing fast!

The typical workflow is to run stamping simulation with ESI PAM-STAMP, bringing the single part to ESI’s Virtual LightRoom for inspection. Images of fringes are generated automatically based on standards defined by Autaza Surface, and AI is used for the detection of defects and their severity measurement. There are built-in standards for severity, and it is also possible to include new standards by training the neural network.

pictures 3 4
Defects on a hood as spotted by inspector match with the ones spotted by AUTAZA’s software.

What role does assembly play in surface defect prediction?

A defect can appear more evidently once parts are assembled. That’s why it’s also very powerful to get the big picture and to be able to simulate surface defects not only on a part but on an entire assembly. ESI’s assembly simulation software delivers reliable results showing the effect of successive assembly steps for all kinds of hot or cold joining techniques: roll hemming, tabletop hemming, spot welding (RSW), laser welding, etc. It predicts surface defects as much as distortion. It helps process engineers find the right parameters to optimize their entire manufacturing process.

Assembled hood in the Virtual Light Room (Courtesy of Ford)

Can you virtualize all steps involved in producing a panel ─ from CAD design to perceived quality assessment?

Absolutely. That is actually the point. ESI’s approach is to start from CAD to enable single part forming simulation, assembly and joining simulation, all the way to painting and heat curing simulation.

When you go all virtual, you can investigate all design options, all sorts of process design without any additional cost, and find the best possible parameters to achieve your targets.

Going all virtual also helps you find the causes of a defect very effectively. Sometimes you may find the die compensation was causing the problem. Sometimes it’s useful to be able to go back as far as to the CAD itself. Correcting a CAD error can be fast and simple if you know how to spot it.

What are the technologies involved in early surface defect prediction?

ESI is a pioneer in material physics. Our DNA is to simulate the way materials behave. But it takes much more than equations and computational physics to virtualize the entire stamping chain down to surface defects prediction! ESI also has extended capabilities in 3D rendering, which are the foundation for our Virtual LightRoom. Along with the prediction capabilities of ESI PAM-STAMP, which uses advanced material models and is able to compute large models such as a side panel with a uniform mesh, and the AI capabilities of Autaza, mentioned earlier, Virtual Reality helps us deliver the best possible solution to aid automotive manufacturers make the right decisions at the right time.

To learn more, download the paper ESI co-presented at the International Automotive Body Congress in Liviona, in 2019: “Quantitative Surface Quality Assessment of Car Outer Panels with a Virtual Light Room” by Ing. Arthur Camanho, Ing. Harald Porzner (ESI Group, USA), Ing. Renan Padovani (AUTAZA, Brazil), and Dr. Carlos Sakuramoto (AEA, Brazil).

Enter your email below and get notified on the latest blog post.