PACE 60-year Anniversary Series: Robotics
In 1961, the world’s first working robot joined the assembly line at General Motors (GM) in New Jersey, USA.
And, true to form in the manufacturing industry, this ground-breaking event happened quietly and without flourish; GM’s executives refrained from publicising their new investment, perhaps afraid it might go terribly wrong.
The robot, called ‘Unimate’, was the brainchild of George Devol and Joseph F. Engelberger of USA-based Unimation Inc.
Essentially a 1,800kg steel arm with a grip at the end, the unit obeyed step-by-step commands stored on a magnetic drum memory.
The Unimate remained on the GM assembly line for the next 10 years, tirelessly dropping red-hot handles and other die castings into pools of cooling liquid; the parts were then moved along to workers for trimming and buffing.
We can only speculate that the 3,000 employees working at the factory in 1961 welcomed the addition of the Unimate, no doubt relieved that their days of handling molten metal and operating dangerous old moulding equipment had come to an end.
Declared a success, in 1967 the first Unimate was adopted abroad, debuting at a Swedish metalworks. Shortly after, the system was launched in Japan, with Kawasaki Heavy Industries – recognising the robot’s labour-saving potential – earning the rights in 1969 to manufacture and market Unimate model 2000s for the Asian market.
It would be another five years before Unimate travelled Down Under. According to the Australian Robot Association (now the Australian Robotics and Automation Association), the first local installation was performed in 1974 by Eric Fender for a client in the automotive industry.
“Although we cannot be sure, we think it was for Ford, performing spot welding,” Engineers Australia National Committee on Mechatronics’ immediate past chair, James Trevelyan, told PACE.
A sheep shearing robot demonstrated at the University of Western Australia in 1980: advances like this led to wide international recognition of Australia’s robotics capabilities. (Image courtesy National Archives of Australia.)
Humble beginnings
By the late 1970s, a number of robot systems were operating in automotive and whitegoods factories around Australia, most performing the more dangerous production tasks of spot and arc welding, machine loading and spray painting.
“Few of these robots had any ‘intelligence’. They were programmed to follow a fixed sequence of movements, with only elementary sensing,” said Trevelyan.
Programmable robotic stacking of pavers onto kiln cars, taken in 2002 at a brick and paver factory in Queensland. This robot could stack any stack-build pattern for a range of pavers and bricks, optimising the kiln firing process. The integrated system included eight robots stacking 12,000 pavers per hour. (Image courtesy of Applied Robotics.)
Though rudimentary, robots grew in popularity both for relieving workers from dangerous duties, and helping raise product quality levels. As demand grew, a number of foreign companies began exporting their systems to Australia, including: ASEA (now ABB) from Sweden; Kuka from Germany; Yaskawa and Fanuc from Japan; and Cincinnati Milacron from the USA.
“Many of these robots were supplied by Australian companies who designed the manufacturing systems, such as Robotics Automation. The first Australian company to actually manufacture robots was Machine Dynamics in Melbourne,” said Trevelyan.
Founded in 1972, Machine Dynamics was originally a manufacturer of pneumatic equipment for the metal and textile industries, but soon became a pioneer in robotic applications, designing and manufacturing systems based on aeronautical engineering principles.
In 1988, the company designed, built and installed a robotic manufacturing line at Ford’s Broadmeadows assembly plant, for the production of the new EA Falcon. The line was the first of its kind in Australia, generating a door every 30 seconds, and remained in production for many years.
It consisted of 13 ASEA spot weld robots and 13 Machine Dynamics Journeyman gantry robots and controllers, all with coordinated movements.
One of the system’s key benefits was reducing tooling changeover time and cost; while previously it had taken five operators an entire shift to re-tool manually, the same task could now be performed in around two minutes by a single operator pushing a button.
A robot cell attends an automated polishing machine at a tapware manufacturer in Sydney in 1995. Brass tap bodies are automatically loaded for final polishing. (Image courtesy of Applied Robotics.)
“The 1980s also saw the demonstrations of robotic sheep shearing at the University of Western Australia and later by a small start-up called Merino Wool Harvesting in Adelaide. These technological advances led to many other robot development projects, and led to wide international recognition for Australian capabilities in robotics,” said Trevelyan.
“Unions realised that without automation and the associated productivity improvements, their jobs would disappear completely.”
In the 1990s, the local mining industry began investing in robotics, following the successful automation of chemical and physical property sampling laboratories in the iron ore sector.
Technology advances
For Machinery Automation & Robotics (MAR) sales manager, Paul Gekas, today’s robots are a far cry from the clunky models he worked with when he first became involved in the industry.
“My interest started when in 1982 as an undergrad at the University of Melbourne I was asked to setup their new electric robot, a Cincinnati Milacron T3 726 (electric robots were just on their way in replacing hydraulic robots),” Gekas recalls.
“I started doing post grad studies at the Uni, but I wanted to get out into industry so I joined Milacron, and afterwards ASEA (ABB). My first field assignment was ‘tuning’ hydraulic robot servo loops at Ford Geelong: you had to trim the PID loop by switching in and out resistors and capacitors of different values; you had about 10 seconds to flick a dozen knobs to the right value before the robot damaged itself or blew a hose… I very quickly developed a dislike for hydraulics!
“My first system build was a little more interesting: one operator and one robot completely assembled the Sunbeam electric irons in Campsie NSW. This was what robots were about: improving the productivity of people.”
An ASEA IRB2000 robot demonstrates its capabilities in the Victorian Pavilion at World Expo 88, held in Brisbane during Australia’s bicentenary celebrations. (Image courtesy of ABB.)
Gekas believes new advances in sensing have been at the core of most major robotic developments, allowing them to work in less-structured environments and deal with much shorter product runs.
“The systems that we’re developing to work on mine sites, abattoirs or atop AGVs [automated guided vehicles] are allowing industrial robots to work in areas that would never have been possible five years ago,” he said.
The next wave
Developing the technology required for robots to successfully work in unstructured environments will be a major focus for robotics engineers moving forward, according to Applied Robotics managing director, Dr Paul Wong, whose interest dates back to 1973 when he published a PhD thesis on Robotics Assembly Using Tactile Sensing.
Wong built Fisher & Paykel New Zealand’s first pick and place robot in 1976, and was employed from 1981 to 1986 as project manager of the Robot Sheep Shearing Programme for the Australian Wool Corporation (AWC), “where our teams developed a robot to safely shear 90% of the sheep's wool – a world first,” he told PACE.
“This robot was one of the first adaptive robots in the world – its movements were not playback, but modified in response to in-built sensors detecting changes in its world.”
According to Wong, playback robots like the Unimate still make up the bulk of units in production today, with most of these working in ‘highly structured’ environments.
“Everything in this environment – the workpiece and the workstation – are precisely known, predictable, and do not change,” he explained.
“Even today’s sophisticated and high-speed processing machines – while being very automated in themselves – still require manual loading at the infeed end. In other words the robots cannot cope with leaning, twisted or flowering stacks, or approximately-placed workpieces on an infeed pallet.”
Engineers in the 1990s first sought to solve this problem by developing sensory technologies that enabled robots to make small adaptations in their pre-programmed moves to account for variations in their environments. Early examples include seam tracking welding robots, vision systems and quality control sensing, and more recently integrated vision and distance sensing systems.
According to Wong, the next step will be omni-sensors that can image the entire workspace, coupled with adaptive and predictive computer modelling to give robots an even greater awareness of their work environment than is possible for human operators. These robots will not only cope with de-stacking an untidy pallet, but also work with ‘unstructured materials’.
“These are the pliable, soft, elastic and simply limp workpieces such as fabrics and soft polymer products, or the low-tolerance assemblies such as wooden crate,” Wong said.
Part of a high-speed servo-gantry to pick and place 30-kilogram bundles of aluminium extrusions, over a nine-metre travel distance, at a cycling time of nine seconds. Taken at a Sydney-based windows and doors manufacturer in 2010. (Image courtesy of Applied Robotics.)
Applied Robotics is already working on making such systems a reality, having signed a contract with the Norwegian Government in 2006 to research and develop Fully Automated Sewing. The first phase of the program – completed in 2008 – saw the successful demonstration of robotic edge matching, and the sewing together of complex sub-assemblies.
“Of course, for robots to perform effective work in these new areas, there will need to be accompanying developments in manipulators such as multi-armed robots, as well as in their end-effectors such as super-dextrous grippers, grippers that can handle limp and porous materials and fragile workpieces,” said Wong.
“The ability to perform real quality control as an embedded capability will also be important, particularly if this function is integrated with the robot’s manipulation functions.”
While Australia may not have started the robot revolution, it has clearly been at the forefront of its evolution since that first Unimate began spot welding on the Ford production line – or, as we like to say, since day ‘bot’.