The Perfect Interaction is Crucial

Robotec Solutions AG from Seon has developed an assembly machine that inserts contact pins into insulators with an accuracy of five hundredths of a millimeter. In the interview, Reto Brumann, Head of Control Engineering, explains what it takes to keep the rejection rate correspondingly low.

You have already realized several assembly automation systems for Fischer Connectors, and now you have been commissioned to build two more systems. What exactly will be automated with these?
The systems press electrical contacts into insulators, thus automating a task in the manufacture of electrical connectors that was previously semi-automatic. While the first systems could only insert one pin diameter into the contact block, these can now press various pin diameters into the insulator. This allows the machines to cover a broader spectrum and assemble even more demanding connector variants. In addition, we were able to further reduce their cycle time.

The Manufacturing Process

Please briefly explain the work process for manufacturing a connector type. What exactly happens there?
The system has several feeders for the unassembled insulators and the various contact pins. From the first feeder, a six-axis articulated robot, guided by a camera, picks an insulator and places it in a holder mounted on a linear table. A second camera checks its positioning, as this is crucial for further processing. Meanwhile, also camera-based, a second six-axis articulated robot picks a contact from the second feeder. This is done according to the assembly sequence, which is why another vision camera first checks if the contact pin meets the correct dimensions. A third camera measures its alignment, as the angle is crucial for correct assembly in the insulator.
A SCARA robot takes over the contact and prepares to insert it into the insulator. Before that, the linear axis moves the insulator under a camera that measures the hole into which the contact is to be inserted. If the pin and hole diameters match, the waiting robot inserts the contact with absolute precision and without play. Then the insulator with the inserted contact moves under a press where it is pressed in force-monitored. Finally, a vision camera checks the pin's angular alignment and also verifies whether it was bent during the pressing process.

With what repeat accuracy do the linear axis and the robots work?
The linear axis has a repeat accuracy of one-thousandth of a millimeter, making it the most precise element in the assembly automation system. The robots position with an accuracy of two-hundredths of a millimeter.

The cadence from pin to pin is seven seconds. What is the limiting factor for a faster cycle time?
The limitation is the precision, which is no longer given at a higher cycle speed. Higher dynamics in the robots lead to overshoots, as a result of which the contacts can no longer be placed precisely. At the latest during the pressing process, they would then be bent.

Importance of Zero Point Calibration

How do robots and grippers know where the contact pin needs to be placed?
There is indeed a hole pattern of the insulator, but its position in space must be defined, otherwise, the insertion will not work. Every position that one of the robots can approach is stored in a coordinate system whose zero point is calibrated across all subsystems of the system. What this means in application is simple to explain. The zero position of the vision camera, which, for example, measures the position of the hole in the insulator, exactly matches the zero position of the robot. If the camera now determines that the hole into which the contact pin is to be inserted deviates by 0.3 millimeters from the x-axis, the robot knows that it must also correct its position by 0.3 millimeters in the x-axis to avoid damaging the contact pin during insertion.

Is this calibration done once, or do all systems realign when setting up a follow-up order?
The system is zeroed by us once. All coordinates that an operator later reads into the system with an order refer to the zero points set by us.

What is the difficulty when a human task is to be automated? When a human inserts a contact pin into an insulator, they can see exactly what they are doing and feel through their fingers if there is any resistance. The robot lacks this sensitivity. It moves into position and places the contact pin. If it is not placed exactly over the bore of the insulator, it will be bent at the latest during insertion, making the component scrap. Therefore, the mentioned calibration is crucial for perfect interaction between robotics and image processing. Of course, it is also important to make the whole thing absolutely user-friendly.

Camera Checks Insulator Position

When the insulator is inserted, is its position checked again?
It could be that it is tilted due to a speck of dust, which could also cause problems when inserting the contact pins. That's right! We do not check how the insulator is placed, but we do check the corresponding bore before inserting the contact pin. If the vision camera detects something in it, the insulator is moved to a blow-off station and then inspected again. If it were tilted as you described, the hole would not be round but an ellipse – and the camera could recognize this to a certain extent.

Modern industrial cameras generate massive data streams. What happens to this data? The images could be recorded for later evaluation, but this is not done in this case. The system is only interested in the position at the moment of image capture, and then the recording is deleted and replaced by the next one.

Logic Distribution on Subsystems

How did you solve the problem of image processing and robotics not overloading the control system?
We distributed the logic across various subsystems. The HMI and the assembly sequence run on a PLC. When a new order comes in, it sends all relevant data to the subsystems. This way, for example, the robot knows what is being produced and independently loads the necessary information from the database to perform the corresponding movements. This approach is also practical in development. It allows the work to be divided very well, so that one engineer can focus exclusively on programming the PLC, the next on the robots, and another on image processing.

Let's talk about the user-friendliness you mentioned. As the manufacturer, you are very familiar with the system's processes. But how do you know that you have developed it to be truly user-friendly?
Before we fully set up or program software, the customer receives a preliminary version of the HMI and the system's operating concept. We then integrate their feedback into the system's operating concept.

The input of a new connector type should be very simple. Is it really that simple, or does it require expert knowledge?
It does not require expert knowledge, but it does require training. For each work step, the user receives an instruction manual that describes exactly how to teach a new component. Depending on the number of poles, 20 to 50 parameters need to be defined, some of which are learned with the support of the vision system. It doesn't require a PhD, but the operator must understand the meaning of the individual parameters.

The assembly system is supposed to assemble hundreds of connector types one day. How do you guarantee this range?
At the beginning of the project, we received the possible insulators and contact pins that are to be mounted on the system. This defined the mechanical area and the design of the robots. On the software side, there are no dependencies on mechanical sizes. However, dependencies arise from the configuration diversity, for example, that contact type 1 cannot be configured with contact type 2. All these criteria must be mapped in the software.

Camera and lighting. What is the idea behind it?
In the vision lab, we check in advance whether the designer's ideas can really be implemented. We work through a list and can already start with the first tests when the assembly reaches a certain stage. This approach avoids errors and accelerates the system's commissioning.

How do you handle the different pin diameters that are to be inserted into the same insulator?
The gripper can handle different pin diameters up to a certain range. If it cannot, the robot independently changes the gripper.

QR Codes for Plausibility Check

In your workshop, I saw dozens of grippers. How do you ensure that the wrong gripper is not accidentally used?
Each tool has a QR code. When setting up a new order, the machine operator scans this when inserting it, so that the control system can perform a plausibility check in the background. If the QR code does not match the information stored in the program, an error message appears.

The feeders for the insulators and contact pins have RGB lighting. Why?
There are two reasons for this. The first reason is related to the image processing system. When it captures the insulators and contacts against a blue background, it can best evaluate the images. The other reason is an optical signal for the machine operator. If everything is okay, the feeder table lights up green; if the gripper cannot find a component, it lights up red.

What should definitely be said about this project in conclusion?
It is an extremely complex system, and I am extremely proud of how everyone put their heart into it and made the project a success. For Robotec Solutions AG, it was one of the most challenging projects in its company history. It took some night shifts to complete the system on time. Thanks to everyone's efforts, we made a perfect landing and created a running system that brings us great joy.

It sounds like you don't want to give up the assembly machine?
(laughs) Yes, that's true.