Rise of the robots

Written by: Steed Webzell | Published:
An illustration of how the ZenRobotics unit fits into a reprocessing system

After a long period promise, will 2016 finally be the recycling industry's 'year of the robot'? Steed Webzell examines the latest technological advances and speaks to some of the companies in the vanguard of this waste management wizardry

Rainer Rehn, chief commercial officer at ZenRobotics, says that this year, more than ever before, will intelligent, multi-tasking robotic systems enter mainstream waste-sorting facilities.

“The robotic revolution is just around the corner,” he says. “Comparable to other process industries, robotic recycling is driven by cost savings generated by unique recognition capabilities made possible using artificial intelligence (AI).”

The ZenRobotics Recycler (ZRR) sorts metal, wood, stone, cardboard and rigid plastic fractions from mixed commercial and demolition waste, and commercial and industrial waste streams, by using ‘smart’ robotics. The core of ZRR, which is available in the UK from Blue Group, is the ZenRobotics ‘Brain’, the software that analyses the data and controls the robots.

The Brain utilises multiple inputs in real time, reacts to changes and even learns from its mistakes. To achieve this it analyses the input from a number of sensor types – metal detector, near infrared spectral camera, a visual range spectral camera and 3D – using advanced AI algorithms. It also applies machine learning to combine the data of two physical sensors to create a so-called virtual sensor.

Learning new tricks

The learning system allows ZRR to recognise and pick more and more materials. Recently, the robot has learned to pick various new objects, such as plastic bags and pipes, while its latest features include hybrid sorting. In short, one ZRR unit can simultaneously sort valuable fractions and reject directly off the conveyor.

Traditionally, waste sorting is based on dimensions; an object’s size, weight or density defines where it lands. In tandem with this, conventional waste processing machines can be classified as little more than fixed-function appliances that are designed for a specific task.

“There have been limited technical alternatives for sorting waste by shape, material or colour, which explains the common practice of manual sorting, as only people have been able to sort objects based on these characteristics,” says Rehn.

However, just like people, intelligent robots can sort multiple materials. Furthermore, thanks to a flexible gripper that opens from 50 to 500mm, the ZRR can handle objects of various shapes and sizes, eliminating the need to resize waste before processing. A single ZRR robot arm can pick up to four different waste fractions with high purity, up to 98%.

High-purity fractions yield better prices and easier sales of sorted materials. The process also creates savings by reducing incineration or landfill needs. Dump less, earn more.

One ZRR can make up to 2,000 picks/hour, lifting various-size objects weighing up to 20kg. A typical installation would deploy two robot arms, making up to 4,000 picks/h, which compares favourably with a human operative who on average achieves around 700 picks/h.

Now multiply that with a three-robot installation that works 24/7 in multiple shifts. That’s a whole lot of sorted waste. Each robot arm has a wide working area (2x2m) and can throw the objects into several waste chutes.

Rehn again: “With time it will become harder to defend investment in energy-intensive, fixed-function technologies and complex processes. Thus, intelligent robotic systems could soon disrupt the waste management industry and override conventional methods.” Blue Group director Eugene Donnelly concurs, adding: “This technology is the future for waste plants as it is almost three times quicker than a human, more accurate and can work over multiple shifts.”

AI is clearly the breakthrough here as it side-steps the previous restriction, namely the highly complex task of teaching machines to sort mixed waste – it’s not possible to tell them everything they need to know.

Deep learning

Emphasising this point is another company working in the vanguard of robotic recycling, Sadako Technologies, which makes use of many different machine learning techniques, including multi-layer neural networks – commonly known as ‘deep learning’.

Of course, this would be impossible if Sadako was not using GPU-accelerated, parallel computing, mainly on the cloud.

The company deploys 2D and 3D cameras to allow machines to recognise objects that they have never seen before. Sadako’s algorithms and real-time parallel software are able to detect and recognise difficult objects in complex environments, allegedly outperforming human capabilities in some cases. For this to be possible, it uses AI algorithms and many terabytes of proprietary labelled and segmented images. The days of an industrial robot being little more than a steel artefact that simply repeats pre-programmed actions are fast coming to a close.

For the recycling industry, Sadako’s Wall-B system is already creating waves, sorting everything from PET and HDPE, through to cans and bricks. It is aimed at flows with a medium presence of valuable materials (between 100 and 500 tonnes per year) where manual recovery work is too costly.

One Wall-B unit can sort 20 pieces per minute, or up to 200 tonnes a year.

From a technical standpoint, Wall-B consists of a high-speed industrial SCARA robotic arm with a compressed air suction grip system that plucks objects from the belt.

In addition, a computer vision system is deployed to recognise valuable material in the images taken by a combination of cameras (patent pending).

An early taker of Wall-B is Ferrovial Services, which says it has installed the first robot equipped with AI that is capable of driving up the collection of recoverable waste materials at treatment plants. Installed in Barcelona earlier this year, the system gathers recoverable PET materials from the many different kinds of plastics moving down the processing line. Recovering PET, which has a high market value, will generate revenue in excess of €50,000 each year. In subsequent development stages at the customer’s site, Wall-B will be programmed to select other types of materials, such as HDPE.

Benefits in bulk

The incinerator bottom ash (IBA) processing facility at Raymond Brown’s A303 Enviropark site in Hampshire, is capable of recycling over 100,000 tonnes of IBA aggregate a year into the construction industry. The IBA, a by-product of incinerating household waste at energy recovery facilities (ERFs), is fed by a front-end loading shovel into a hopper with a variable speed belt feeder, before being passed under an over-band magnet to recover large ferrous metals.

The IBA is then fed into a 23m-long, rotating trommel screen. Here, all the smaller material (less than 65mm) passes through the screens while the larger material is delivered to a picking station for further recycling.

The smaller material is fed onto a second incline conveyor and under a secondary over-band magnet, which recovers any smaller ferrous metals, before delivery into the screen house and on to a large binder screen, where it is split into three sizes.

All the screened material is delivered via respective conveyors over magnetic head drums to recover any residual magnetic materials. The three grades of material are then passed over a concentric eddy current separator to recover light and heavy non-ferrous metals. Finally, the three remaining grades of aggregate material are sent to a radial collection conveyor where the blended material becomes a fully processed IBA aggregate, which, e.g. is used as a sub-base material for road construction.

Maintaining the theme of bulk-handling aggregate, Tesab reports that it has installed its latest generation Trackstack 6542TPS tracked mobile picking station at a customer in France.

Based on the design of the existing 6542, the new-generation 6542TPS is designed for greater capacity, lower operating costs and quicker set-up time. It includes a direct feed hopper that can be fed by wheeled loader or excavator. Furthermore, the system houses a cast-iron pump and internal drive gearbox at the conveyor head that is said to provide 3.5 times more torque than standard set-ups.

This material is protected by MA Business Ltd copyright.
See Terms and Conditions.


Please view our Terms and Conditions before leaving a comment.