Binder jetting printer with a camera above the sand bed, a technician with a tablet compares the printed layer with the CAD model

How ExOne detects layer defects in 3D printing before the next layer

ExOne builds binder jetting systems for industrial 3D printing. A camera compares every printed layer with the CAD model, and an RL agent adjusts binder flow and recoater speed in real time.

Discuss your project

Customer: ExOne GmbH
Industry: Industrial 3D printing (binder jetting)
Services: IoT Business Model Builder, app development, AI Operator
Technology: Computer vision, reinforcement learning, MindSphere

Starting point

ExOne builds binder jetting systems for industrial 3D printing with sand and metal. The machines run worldwide, for example in foundries.

The S-Max Pro series already had everything needed to make machine data usable in the cloud. But the data was not yet being used. ExOne wanted to turn automation, software and AI into business models that deliver measurable value to customers and users.

Then there is the process itself: the part is built layer by layer. A defect in one layer carries over into all the following ones.

Implementation

It started with our IoT Business Model Builder. ExOne’s executive management, sales and development teams evaluated use cases and business models together with customers. The result was a basis for a management decision.

This led to ExOne Scout, the app for the machine, based on Siemens MindSphere. It shows process data in real time, also remotely, and reports the print status via notifications. Scout runs on the web, iOS and Android and is in use at customers in six languages.

With the AI Operator, ExOne is taking the next step. A camera captures every printed layer and compares it with the CAD model in the build space. An RL agent detects layer defects before the next layer and adjusts binder flow and recoater speed in real time.

Target/actual comparison of CAD and build space in every layer
Reinforcement learning for binder flow and recoater speed
Edge agent directly at the machine

Closed loop at the printer: a camera captures every sand layer, an edge device passes the image to an AI model that adjusts binder and recoater

Dashboard with inspected layers, one was corrected during printing, next to it rising bars and a smartphone with a status message

Result

Layer defects are detected in the layer in which they occur. The print is corrected during the process or stopped in time, before scrap is produced.

Completed print jobs can be analyzed and compared, which results in recommendations for action. Because it is known at all times whether an S-Max Pro is printing, manufacturing networks share their capacity better.