Case study: Robots operate older press brake including keyboard
A research project at the Deggendorf Technical University(Center for Applied Research at the Cham Technology Campus) has now progressed to the point where the approach being pursued can be described as successfully implementable. Under the direction of Peter Lingauer (Chair of Prof. Dr.-Ing. Wolfgang Aumer), an older press brake (here built in 2006), which does not have any interfaces, is to be fully automated. However, the automation is only intended for larger series. The background is the well-known one: A medium-sized company in the region lacks skilled workers and therefore cannot process all orders; it misses out on sales. Hence the objective that the increasingly scarce skilled workers should process smaller series and more complex orders per day. The monotonous larger series, on the other hand, should be created automatically during the night. The challenge: It is not enough to simply place sheet metal in the press, the machine must also be operated occasionally via its touchscreen. For this, it must be read. However, an interface is not possible because of the age of the machine. Although the press is already 16 years old, it can continue to operate for another few decades. A replacement was therefore not an option.
Funding made the project possible in the first place
Those involved are fortunate that the project is being funded by the Free State of Bavaria. The Free State of Bavaria is generally generous in its support of future technologies (at least in comparison to other German states) and is also happy to support rural areas. Because if it offers interesting jobs and learning opportunities (regional universities), the rural exodus is slowed down and the rural area remains livable. Without this support, the whole project would definitely not have been undertaken, as the risks were considerable. The medium-sized company, which still bears a considerable part of the costs, receives the rights and can subsequently ensure scaling. For example, by passing on the know-how gained to a robotics company in return for license fees.
Challenge of human-robot alternating operation and existing layout
As with any company that has grown rapidly, the space available is limited. Over time, one machine has "grown up" to the next and there is hardly any space. Since an employee is to continue operating the bending machine during the day, the automation solution must not be stationary. It must not only be mobile in the narrow sense, but also quickly mobile. Longer changeover times would of course be a knockout criterion. In total, two robots are used. One places the metal parts in the press brake. This is mounted on a trolley and can be moved away quickly. A mark is placed on the floor so that the mobile station is quickly back in the right place the following day. The other robot is used to communicate commands to the machine. This is attached to the machine (see photo above) with profiles, as we know them from item, among others. This does not need to be dismantled for the day shift. The relevant machine screen can be turned away instead. In the evening it is then approximately turned back to the robotics position. (The related software - see below - automatically detects the screen, so no calibration is required).
The basic criteria for the "workhorse", i.e. the robot that places the metal parts in the press brake, was a reach of at least 1.20 m and a payload of at least 10 kg. However, the simulation then showed that a typical six-axis cobot would reach its limits. Reason: In order to be able to work permanently at night, the cobot must be surrounded by grid boxes with material or to pick up the bent material. This in turn requires a high degree of maneuverability. The KR 1410 from Kassow Robots best meets these criteria, and is also remarkably fast.
The selection of the small robot that is to operate the keyboard or the touchscreen was more difficult. The criteria here were a range of at least 30 cm, a weight of well under 10 kg (which is attached to the arm of the machine) and good repeatability. The Cobotta from Denso would have met all these criteria, but with a price of around 15,000 euros it would also have exceeded the budget. However, the job is not so demanding that it is absolutely necessary. After much research, the decision was finally made in favor of the Ned 2 from Niryo. At 4,000 euros, the price of this French miniature robot is considerably higher than that of Chinese models. But after its predecessor had already been sold more than 3,000 times, the company had greater confidence in it. There is no question that if a company should ever want to automate several bending machines, an endurance test of the Chinese models should be seriously considered.
As of today, both robots - Kassow Robots and Niryo - have proven to be absolutely the right choice.
Software and hardware
A standard Sensopart camera and software are used to handle the workpieces. The basic prerequisite for the entire automation, on the other hand, is the Opdra software, for which the author is largely responsible(contact him). (automatica visitors will probably know her from the Neura Robotics booth, where she operated a kitchen machine ).
Opdra reads out a screen/display purely optically. The relevant can be marked easily. High image sequences are possible. Likewise the depositing of scripts ("if this stands, the robot should do this").
In this case, a simple USB Logitech camera was mounted on the arm of the Niryo so that the screen of the sheet metal bending machine can be read. Using an attachment from the 3D printer, the Niryo then makes the necessary inputs. The input is again monitored by means of the camera. Any errors in the input (it rarely happens that a wrong field is touched) are then corrected.
Of course, there was a lot more to consider. For example, the thickness of the sheet metal parts used must be checked. However, all this is not to be considered critical, but nevertheless requires work input for the first time. In return, the company can run at least two more shifts (nights, weekends) and thus increase its sales.
The young Opdra software thus has another use case. At the beginning there was the operation of a measuring machine by means of a Franka Emika. The corresponding video is very informative(link). Using the example of a Universal Robot, it was shown that colors can also be recognized. In another case, a Kuka industrial robot is used to write to a database.
Ultimately, the approach is transferable to virtually all cases in which a robot is to receive information from a machine, provided that the machine has a screen. This applies to both industry and the laboratory. And this is even the case when an interface would be possible. Because this often costs a lot of money and is inflexible.
The business model of VisCheck GmbH, the developer of Opdra, consists primarily in the further development and distribution of the software. The concrete implementation on site is gladly left to the integrators, who can fall back on support if necessary. Previous partners have been able to get along with the software. For integrators as well as OEMs, the software is interesting for two reasons:
- Automations are feasible that were not before. In other words, software generates additional revenue.
- Customization at the customer's site adds value. On the one hand for simple applications and then often through complementary optical programming (pattern recognition, quality control).
As mentioned, the author is available to answer questions. He is the managing director of VisCheck GmbH.