Machine Vision in IIoT
How machine vision technologies help to overcome new challenges related to connected and automated production.
Industrial companies are confronted with several new trends that will fundamentally change production and logistics processes. For example, the term “Industry 4.0,” which was coined in Germany, stands for the digital networking of people, objects, and systems to create integrated production processes. In international jargon, it is referred to as the Industrial Internet of Things (IIoT). All technologies, systems, and components that are involved in the industrial value creation process are connected to each other as well as to company networks and the internet. Smart factory is another trend that forms a part of the IIoT development. It refers to a production environment in which machines, installations, and logistics systems interact independently, without human input, and partially communicate through the IIoT. These automation trends are still fairly new, and the relevant technologies and systems will be steadily developed further in the future.
Robotics plays a special role in IIoT and smart factory scenarios. The recently published 2015 World Robot Statistics from the International Federation of Robotics (IFR) reflects the current developments: it expects that around 1.3 million industrial robots will be in use worldwide by 2018. The current global market value of robot systems is USD 32 billion across industries. Robotics is making strong inroads in the automotive industry, where investments in this technology rose by 43 percent from 2013 to 2014. In the manufacturing sector, there are on average 66 robotic units per 10,000 workers worldwide. South Korea is the world leader in automating processes using industrial robots, followed by Japan and Germany. Surprisingly, according to the study, the U.S. is only in seventh and China is in 28th place.
New, compact generation of robots
A new trend is also visible within the robotics segment itself, as a new generation of robots is taking the automation and manufacturing industry by storm. These industrial robots are much smaller, lighter, and more compact. They operate with great flexibility and work closely with their human colleagues. They often feature one or two arms and sometimes even a head. In addition, they are cheaper than large stationary five-axis robots and can also be used as mobile units owing to their light weight. Another special feature: the robots can complete alternating tasks, offer better support to their human colleagues, and also take over some of their work tasks if that becomes necessary, for example in the case of illness.
The ability to exploit the advantages of the new production trends requires well-thought-out accompanying technologies that support the further development of automation. This also includes machine vision, which has grown by leaps and bounds in the past few years. The technology uses state-of-the-art image acquisition devices, such as high-resolution cameras and sensors. They are used to record the production processes—frequently from different angles—and generate digital image data that is processed with the help of machine vision software. In this way, production processes can be visually monitored, while weak spots and optimization potentials can be identified. The technology is also extremely fast: algorithms process the digital image data in milliseconds, thus paving the way for real-time applications. In addition, machine vision also impresses with robust identification processes and a high detection rate.
Quick and flexible set-up of robots
The technology plays an important role particularly in the interaction with the new generation of robots. The majority of the new industrial robots already come with one or more cameras and integrated machine vision functions. Preparing the robots for different application scenarios should be rapid and flexible—without long training of the various tasks and without cumbersome set-up processes. Similarly, machine vision applications should also be easy to create, using certain software, for example. Its central element is an image-centric user interface, which intuitively guides the user through the application. In contrast to conventional programming tools, users do not have to worry about dealing with complex codes, command lines or parameter lists, but rather benefit from an easy-to-read visual display, similar to a WYSIWYG (what you see is what you get) editor.
In addition, the software also contains a collection of standard vision tools such as image acquisition, calibration, alignment, measurement, counting, checking, reading, position determination, and defect detection. Using an integrated feature, objects that must be identified can be detected, highlighted, and selected with one click by simply moving the mouse cursor over an image. This does away with the cumbersome configuration of complex parameters, thus saving time and costs during the development process. Now, comprehensive machine vision applications can be created quickly and easily, without the need for in-depth programming experience or image processing knowledge. This benefits all industrial sectors that require new applications due to quickly changing requirements in production chains. It is no longer necessary to write a completely new program for each new task.
Exact positioning of workpieces
Machine vision processes optimize a variety of links in the industrial value chain. For example, the technology reliably detects objects in production processes, safely identifies the position of workpieces, and finds the optimum alignment for the same. As a result, robots are able to accurately grasp and process them, which increases safety in automated production processes. Two-dimensional processes have so far been the standard method in this area. Typically they are used to determine the position of horizontally moving objects (on a conveyor belt, for example), but they cannot identify three-dimensionally acting objects such as interacting robots. Therefore the suitability of 2-D methods for use in highly-automated production scenarios is limited.
Such environments do better with three-dimensional machine vision technologies such as “3D scene flow.” In contrast to conventional methods, “3D scene flow” uses a multiple camera setup: several cameras positioned at different locations view production processes from different angles. This creates a three-dimensional movement profile, which can be used to identify not just the exact position of objects, but also their movements in the three-dimensional space and their speed. The 3-D technology is particularly interesting for robot-supported, highly-automated production processes, since it makes the collaboration between humans and machines more efficient and also safer. For example, it can precisely determine the movement direction of mobile robots that are moving independently through production buildings, preventing dangerous collisions with humans or vehicles.
Reducing expensive machine shutdowns
The work carried out by stationary and large five-axis robots used for welding and other assembly processes can also be optimized: in the automotive industry, these robots operate in separate areas to which humans do not have access. If a worker nevertheless steps over a certain line, light beams or sensors will stop the robot to ensure that the “intruder” is not hurt. Valuable time goes by until the robot has started up again, which leads to expensive interruptions in production processes. These processes can now be made safer and more efficient using the new 3-D technology. The three-dimensional movement profile precisely identifies the action radius of the robot, so that imminent collisions with humans can be detected early on. This does more than simply increase safety. The company also saves costs, since the frequency of robot shutdowns can be reduced with the precise, three-dimensional monitoring of production processes.
Modern machine vision technologies not only assist production processes, but also help to optimize quality assurance processes across industries. For example, they can be used for the exact scanning and inspection of tool surfaces in the metal industry. Now defective parts can be safely identified and automatically rejected before entering downstream process chains. The high speed of the image processing systems also allows for quick and automated inspections of large batch sizes. In the electronics industry, the technology can help to safely identify and detect errors for many different components and electronic parts. Also in packaging processes, machine vision can play an important role: the quantity of products in packages can now be reliably determined, ensuring completeness.
Seamless communication between products and manufacturing equipment
Finally, machine vision methods also optimize logistics and transfer processes, particularly when it comes to the communication between products and manufacturing equipment in modern IIoT scenarios. Here, products contain all production information in machine-readable form, such as bar codes, QR codes, or color coding. Modern image processing systems can read this coded information reliably—even when the codes are defective (overexposure, too narrow, wide or blurry code bars). The data thus read out can be used to automatically control the product’s path through the production equipment and individual process steps. V&S