Interest in bin picking has been existent for years. A feeding solution that is compact, quiet and flexible would benefit any forging plant. The primary reasons are maintenance and cost. Commonly used vibration feeders can be quite expensive and require intensive maintenance programs during their service life. In comparison, bin-picking robots with integrated-vision systems can do the same job at lower cost and with little maintenance. For example, the annual maintenance costs of mechanical feeders can be as large as the initial acquisition cost of a 3D-vision system.

Vibratory feeders are tried and true in the industry, however, and bin-picking robots have only recently begun to find application in forging plants. We have been helping forging plants automate bin picking since 2016 and have discovered the most common mistakes when automating billet feeding with 3D-vision-guided robots.

Don’t Start Bin Picking with Complex Shapes

Selecting forged parts with complex shapes as an application for your first bin-picking project is probably not a good idea. Working with such parts requires a deeper understanding of how 3D vision works, more extensive programming of the robot, sophisticated grippers and longer integration timelines. That is why it is better to start with simple geometries, like cylinders, to succeed quickly and gain experience.

Don’t think twice about it. There is one perfect application in every single forging plant with which to begin your bin-picking expertise – feeding cylindrical billets to the induction furnace straight from the bin. This is an industry-proven application.

Get the Hardware Design Correct from the Start

Many people think the success of a bin-picking application depends mostly on the vision system (i.e., the quality of the 3D camera images and the software’s capabilities). Indeed, the vision system must find all visible billets to ensure that there is always a next billet that is reachable for picking. However, correct detections are only 50% of a successful application. The other 50% is about the hardware. Educating yourself about hardware and cell design will help secure the bin-picking cell’s reliability in advance.

Billets can be lying in many different orientations and hard-to-reach corners or sides of the bin. That requires an adaptable gripper that can pick billets from many different directions deep into the bin. The gripper and the robot must have the ability to deal with occasional collisions with other billets in the bin. Humans collide with the environment all the time. A collision is not a problem if the hardware is designed for it and if the steady flow of billets to the oven is not interrupted because of that.

Furthermore, a simple but effective game-changer for obtaining a truly reliable bin-picking cell is a re-pick station. After picking a billet from the bin, the robot places the billet in a kind of fixture to obtain an exact picking position and orientation before re-picking the billet and bringing it to the oven.

Examples of the re-pick station include a V-shape slider using gravity or a piston to obtain alignment. Using a re-pick station maximizes robustness at the cost of 2 seconds of cycle time. It guarantees perfect alignment of the billet in the gripper, resulting in a robust grip and accurate placement. It allows you to add a quality-control check verifying the billet dimensions (with lasers) as a bonus.

Some Questions Answered

For operations new to automated bin-picking and vision systems, it is appropriate to ask a few questions early in the design process. However, knowing what to expect from a new bin-picking system is difficult for those without experience and before you put the system into production at your plant. Based on our experience over a large number of actual installations, here are the answers to some questions from real bin-picking production cases.

Are there cycle-time limitations?

One of the first questions we ask our customers is what cycle time they wish to achieve. If, for example, a customer says they require a steady feeding flow of 6 seconds per billet, this would be too short of a cycle time for bin-picking applications. The bin-picking cycle time is typically in the order of 12-15 seconds with one robot. This cycle time must be faster than the oven cycle time. That way you can restore your safety buffer, which typically contains about 10 billets.

If optimizing the cycle time is critical, we consider installing two robots and designing the setup to minimize robot motions outside of the bin. In many cases, the oven is slow enough and you don’t need extra optimizations. The table shows a cycle-time example from one of our customers. In this case, the conveyor and the oven are located much higher than the robot and the bin, which results in 6-8 seconds of extra cycle time. However, the induction oven dictates a cycle time of 30 seconds. So, in this case, it is OK to have a robot doing extra motions to reach the conveyor belt.

Are special bins needed?

In our experience, the best results you can get are from a bin that is always placed in a defined and fixed position. It could even be slanted to facilitate robot-arm picking. Bin walls can also impact performance – straight or even slanted walls help improve the application’s robustness. Also, if you have bins that have curved edges or have been deformed through wear and tear, you do not have to replace them. Worn bins can be used as intermediary bins to transfer billets to the one fixed straight bin used for bin picking.

Which types of billets work with bin-picking systems?

Shape, diameter and length are important factors in bin-picking performance. At Pickit, our system supports round billets more than 30 mm
(1.2 inches) in diameter. For other geometric configurations, such as thin billets or round-cornered-square billets, we advise having an integrator to set up Pickit for your precise situation.

Is the bin always completely emptied?

In a typical production situation, it is possible that the last 5-10 billets could be stuck in the corner and not be picked. To make sure that the required number of the billets goes in the oven, load the last bin of the production run with extra billets. That way you can prevent operators having to intervene before the end of the process.

Conclusion

A robot with an integrated 3D-vision system that gently picks billets from a bin and feeds them to the oven can be a cost-efficient and compact alternative to vibration feeders. Knowing how to tackle bin-picking challenges will help you reach the desired reliability and unlock the benefits of robotic billet handling.