TractorsSkyHawk: Auto­mated quality assur­ance at the Mannheim plant

All trac­tors supplied by the John Deere factory in Mannheim meet the highest quality stan­dards. An auto­mated camera robot now helps employees to detect and correct errors in the produc­tion process.

The John Deere plant in Mannheim is bustling with activity. Trac­tors roll along a conveyor belt which winds its way through the entire produc­tion hall, from one assembly station to the next. With each step, they progress from a frame to a complete tractor with a cab, bonnet, tyres and lights. The conveyor belt now takes the trac­tors through a large archway halfway through their journey. Blue lights on the floor mark the start of the working area of one of the latest achieve­ments in quality assur­ance – the SkyHawk.

The heroic name is given to a high-tech camera system mounted on a six-axis robot. It thor­oughly checks the trac­tors before they are fitted with cabs, bonnets and other compo­nents at the next stations.

A keen eye

The SkyHawk camera scans 50 inspec­tion points within three minutes and compares the condi­tion of the tractor against stored images of the target construc­tion.

Pursuit of perfec­tion

The highest possible tractor quality is one of the most impor­tant goals for all employees at the Mannheim plant. Linus Baumhauer, plant manager at Mannheim, explains: “With the SkyHawk project, we are bringing the basic idea of the See-and-Spray tech­nology devel­oped for our customers into our own produc­tion. While farmers are concerned with the sustain­able and targeted appli­ca­tion of crop protec­tion chem­i­cals in the field, SkyHawk helps us detect and avoid devi­a­tions in produc­tion at an early stage.

SkyHawk helps us detect and avoid devi­a­tions in produc­tion at an early stage.

Linus Baumhauer

The aim of the project is to use a robot to repro­duce human move­ment sequences and to recog­nise and inter­pret assembly results using indus­trial image processing. This enables us to reduce our employees’ work­load while auto­mat­i­cally moni­toring the quality of our prod­ucts.”

Linus Baumhauer, plant manager at Mannheim

Tobias Trunk, factory automa­tion lead at Mannheim, is respon­sible for the concept and design of the SkyHawk system. Together with his team, he first had to find a solu­tion for seam­lessly inte­grating the inspec­tion system into the factory struc­ture in Mannheim, which had grown over the years. “There are only a few compa­rable systems in the world,” he says. “And they are all set up parallel to the assembly line – this reduces complexity because the robot can then move the camera more easily to all parts of the vehicle. However, this type of assembly was not possible for us because the space only allows for a gate system which works across the manu­fac­turing line above the trac­tors.”

So the camera can inspect every corner of the tractor, it is mounted on a six-axis robot which moves back and forth on a linear rail above the tractor.

To find out how this system could work, Tobias and his team first built a fully digital proto­type of it, which they used to thor­oughly test the concept. The second step was to set up the system in an empty hall on the Mannheim factory premises. “We had to over­come a few hurdles, espe­cially before the hard­ware and soft­ware could really work together. After all, we have to be able to map and test a wide range of para­me­ters,” says Tobias.

To enable the camera to inspect every corner of the tractor, it is mounted on a six-axis robot that moves back and forth on a linear rail above the tractor.

To find out how the system could work, Trunk and his team first built a fully digital proto­type of the system, which they used to thor­oughly test the concept.

Construc­tion with a dead­line

The route that the camera takes around each tractor model is precisely programmed by the experts, as are the inspec­tion points at which the camera has to stop briefly. The foun­da­tion for this is digital models of the trac­tors: In live oper­a­tion, the camera does not fly in front and over the phys­ical tractor, but over a digital “coat” which covers the machine. SkyHawk receives infor­ma­tion about which model is to be tested and in which config­u­ra­tion via a radio frequency iden­ti­fi­ca­tion (RFID) chip in the tractor’s fuel cap, which carries this infor­ma­tion.

After the team had put in many hours of testing and programmed many lines of code, SkyHawk was ready for use in live produc­tion. A fixed time window of 14 days during the company holiday at the Mannheim plant was avail­able for its instal­la­tion. During this time, the team disman­tled the SkyHawk system in the test hall and set it up in the tractor produc­tion facility. “It was a very intense time for all of us,” recalls Tobias. “But because we all worked together as a team really well, the system was up and running before produc­tion started again after the shut­down.”

SkyHawk has landed

The SkyHawk system was installed and put into oper­a­tion during the 14-day company holiday at the Mannheim plant.

Team­work between humans and machines

In the first few hours and days of SkyHawk’s deploy­ment, Tobias and his team were constantly on hand to quickly remedy any start-up diffi­cul­ties and elim­i­nate any bugs. The system has now been in use for several months. It has proven itself, and inter­ac­tion with human colleagues has also become estab­lished. In each produc­tion area, there is now a perma­nent “stand-in” who wears a smart­watch on their wrist. As soon as SkyHawk detects an irreg­u­larity, the system analyses in real time which produc­tion area can correct the problem. The person in the field then receives an alert on their smart­watch, which indi­cates the noted irreg­u­larity. This allows them to go directly to the SkyHawk station and fix the problem before it causes further issues.

The core team behind SkyHawk: Tobias Trunk, Timo Niebel and Bryan Horr.

Tobias is satis­fied: “The SkyHawk project shows how we can use and inte­grate tech­nical systems into our produc­tion which further improve the quality of our trac­tors while supporting workers’ tasks here at the plant. This is how useful automa­tion works.”