Case Study: Custom Drone Production

VIUN
5 min readDec 10, 2021
A factory floor with a conveyor belt where VUE’s system is inspecting a drone.

The Challenge: Visual Inspection of Custom Drones

The Swiss Smart Factory (SSF) is a branch of the Swiss Innovation Park in Biel, Switzerland. Their goal is to promote and showcase new and innovative industry 4.0 technologies. To do so, they built their own production facility in the heart of Biel, where they 3D print custom drones. Visitors can easily customize drones to their preferred color, length, and number of arms, which will then be printed and manually assembled in the SSF’s facility.

3D printing technology allows the SSF to manufacture various parts and iterate quickly yet continuously to improve the design and functionalities of their drones. However, the added customization makes the manual drones’ assembly more complex and error-prone, even with the help of interactive and guided assembly workbenches.

Close-up of a drone on a workbench getting assembled.
Guided assembly of a customized drone in the Swiss Smart Factory.

Thus, the SSF asked VU Engineering to use their Autonomous Visual Inspection System to provide a rapid feedback on the drones’ assembly by comparing a picture of the assembled drone with its corresponding customer’s order.

The Solution: Autonomous Visual Inspection System

It’s difficult to use traditional rule-based machine vision applications when inspecting customized drones with their ever-changing colors, shapes, and arrangements. An added challenge was that VUE’s inspection machine had to be movable along the SSF’s conveyor-belt to adapt to the continuous improvements and restructuring of their manufacturing processes. For this, VU Engineering decided to integrate a camera and a small computer onto a mobile light source. This lightweight setup proved very mobile and occupies very little space on the manufacturing floor.

The customizable parts weren’t the only variable through the process that VUE took into account. The conveyor belt was situated next to a large window, allowing the sun to shine directly onto the inspected parts and potentially disrupt data capture. However, VUE’s Autonomous Visual Inspection System can work independently of the camera type and adapt to open inspection environments with changing light conditions and parts position.

A factory floor with a conveyor belt where VUE’s system is inspecting a drone.
VU Engineering’s lightweight setup to visually inspect the assembly of the customized 3D-printed drones.

The Connector

The mobile and light-weight inspection machine is equipped with VUE’s connector, which connects the machine on the factory floor to the system processing the images to perform the visual inspection. It also connects to a tablet to let the operators interact with the inspection process and the inspection machine.

VUE’s connector can link to inspection machines through various communication channels, making it easy for clients to integrate it into their existing systems. The connector is responsible for processing the data itself or for sending it to another machine, either in the cloud or on premise. It also hosts the interface for the human operators, letting them easily connect and interact with the machine and the inspection process.

The User Interface

Core to VUE’s product is the ease-of-use of its systems. With the tablet’s simple tactile user-interface, the operators can easily draw and select different options to correct the inspection results proposed by the machine or add new information and parts to inspect.

With just the tablet, the operators can manually start inspections, review the inspection results, and teach the system. VUE’s system does the heavy lifting so that its clients and their operators can focus on their existing tasks.

A screenshot of a user interface with a picture showing the inspection results and different buttons to modify or add information.
User interface used by the operators to correct the detections of the visual inspection pipeline and add new inspection points.

The Autonomous Visual Inspection System

VUE’s Autonomous Visual Inspection System (AVIS) is the system in charge of collecting, organizing, and processing all the data generated and required to do the visual inspection. It is composed of three independent components that operate together. First, the visual inspection pipeline processes the images coming from the inspection machine and returns the inspection results. The second component is the learning system that generates the inspection pipeline and improves its different components according to the new information given by the operators (e.g. new defects or parts to inspect). Finally, the data collected and generated during the inspection process is stored. This is used by the learning system to assess performances and continuously improve the visual inspection pipeline. It can also be integrated with other tools used on the factory floor that might be required for compliance or traceability.

In the specific use-case of the SSF, the decision was made to have those three components hosted on Microsoft Azure Cloud. Having the full system in the cloud greatly reduces the cost of the ownership of the solution. Indeed, the SSF doesn’t need to own expensive hardware necessary for the system to function. In addition, the infrastructure can easily be scaled up or down according to the amount of inspections that needs to be done. However, as each of the components is independent, they can be installed on different systems at different locations depending on the use-case.

The Results: Faster Responses on Assembly Errors

In order to inspect the SSF’s 3D-printed drones and showcase the learning capabilities of VU Engineering’s AVIS, it was decided to create a demonstration focusing on a more specific use-case happening at the SSF. As visitors could select a different color for each arm of their drone, it was decided to introduce a new pink color that was not available at the initial deployment of the system. Visitors could witness how VUE’s AVIS would react when inspecting a drone with a pink arm, and how it is possible to teach the system to recognize it.

The building of the customized inspection system took less than two weeks and was easily deployed in only half a day, allowing the Swiss Smart Factory to rapidly have a working solution on their conveyor belt to swiftly respond to assembly errors.

Follow VU Engineering on LinkedIn and YouTube to see more and stay updated on our new use-cases.

Want to improve the quality of your product and reduce your inspection costs? Contact us below!

--

--