Factory automation has played a pivotal role in the evolution of manufacturing from the early days of manual control through relays and switches to the advanced digital platforms that are seen today. Increasing productivity, achieving mass production, and improving efficiencies have been a perennial quest for manufacturers and have been catalysts for advancements in automation. We are at a crossroads once again as a new and vastly different era of automation dawns. In the 20th century, the principal driver of factory automation was mass production; in the 21st century, the driver will be manufacturers’ need to manage rapid innovation cycles, complex personalization requirements, and intensifying cost pressures in a global landscape with stiff competition.
Many manufacturers that have enjoyed leading market positions with their legacy machines and production processes will soon find their value proposition either rendered obsolete or, at the very least, challenged by new competitors with innovative ideas. This demands a reimagining of factory automation: a new architecture that can meet the complexities and demands of the future.
For the industry at large, it is imperative to realize that this will be an immediate, tectonic shift rather than the gradual evolution seen in decades past. It will require a dramatic change in the way we perceive factory automation. Many of the rules are likely to be rewritten.
The future of automation will be completely consumer-centric
A changing B2C environment necessitates or compels for a change in B2B space. Manufacturers, who cater to B2C, need to innovate and adapt themselves to the changing dynamics of the B2C world. Digitalization is the means for the industry to deal with upcoming changes.
In the pursuit of digital industries, we are poised to see a major disruption happen in the world of industrial automation. Reinventing factory automation, which lies at the heart of industries, will be a prerequisite for achieving digital industries.
Artificial Intelligence is the lynchpin of the Future of Automation vision
Its application in the automation space will convert passive assets into self-learning objects that can continuously improve and enable automated process optimization.
Manufacturing has taken a definite lead when it comes to envisioning and implementing AI for industrial automation. Business considerations around efficiency, cost, regulatory compliance, security, and risk make AI a perfect solution for addressing the changing demands of industrial automation. Functions within automation that are largely manual will become hot spots for AI. The technology is already being implemented in most modern factories in some form. With traditional automation approaching a saturated state, the dawn of AI will unleash a new breath of life onto factory shop floors and make automation more plausible in the age of digitalization.
The best way of quality assurance does not have to be sexy
In manufacturing processes, quality control can cause a bottleneck that costs time and resources.
If AI technology is placed in an automated manufacturing process, the neural network can learn what distinguishes an error pattern, even during operation.
As our AI expert Ingo Thon says from AI’s point of view quality control is not sexy but very reasonable in the implementation because there are a lot of benefits for our customers production. But how does reasonable AI work in production?
The quality assurance of components in production is a complex process now and errors are often not or incorrectly detected. This leads to the fact that it is no longer possible to produce economically and ecologically.
Our quality inspection approach shows the detection of defects in components before the final assembly of the product. For this purpose, the individual parts are inspected in an optical procedure. Defects often occur here because screws are missing in the areas provided for them in the product.
Artificial intelligence can bring an enormous improvement. Because if it is possible to recognize the defect with the human eye, it can also be learned by artificial intelligence.
The process can be described in 4 steps:
First, we need data in the form of images. This means that the defective components must be imaged from all perspectives and in good quality. The better the input of the information, the better the AI can be trained.
The second step is to label the images. Labeling is a process of telling the AI exactly what is on the picture e.g. error: “missing screw”.
In the third step we let the neural network learn with this information. The more images with similar errors were labelled, the better the neural network can be trained to recognize these errors.
In the final step, we now want to integrate the trained artificial intelligence into the quality control process in the plant. For this, we deploy the neural network to the TM NPU, the technology module for the Simatic S7-1500 that can process neural networks. Now defective parts that are guided through this process and optically detected by a camera can be immediately detected and rejected from the process.
Whereas the quality assurance of complex products was previously only possible manually, the inspection can now be automated and thus made faster, more cost-effective and more reliable by implementing the appropriate AI.