Artificial Intelligence & Industry 4.0
5 Manufacturing Applications for AI
If you made a list of the most overused buzzwords in manufacturing today, ‘artificial intelligence’ (AI) and ‘Industry 4.0’ would be right at the top. (Runners-up would include ‘IoT’, ‘Smart Factories’, and ‘Cyber-Physical Systems’, with an honorable mention for ‘Blockchain’.) It’s unfortunate, because these are more than just buzzwords, and understanding the concepts behind them is crucial to staying competitive in modern manufacturing.
A recent Forbes survey on AI found that 44% of respondents from automotive manufacturing classified AI as “highly important” to production over the next five years, while 49% said it was “absolutely critical to success.” And yet, more than half of the respondents (56%) from automotive said they plan to increase AI spending by less than 10%.
Meanwhile, TrendForce has predicted that the global “smart manufacturing” market will be worth $320 billion by 2020, with a compound annual growth rate of 12.5%, and PwC reports that analytics and AI-driven process/quality optimization will drive an expected 31% increase in smart factories over the next five years.
What is going on?
Not long ago, the manufacturing sector was swept up along with everyone else in the early days of the current AI spring. Then came the predictable skeptical backlash: comparisons of artificial intelligence to snake oil, claims that “’AI-powered’ is tech’s meaningless equivalent of ‘all natural’. As a result, manufacturers may be experiencing a sort of conceptual whiplash.
Is AI a manufacturing cure-all or a quack remedy?
It should come as no surprise that we see it as neither; artificial intelligence is a tool, and like any tool its usefulness depends on the context in which it’s being applied. That’s why we created this list of the most promising manufacturing applications for artificial intelligence today.
1) Predictive Analytics
We might as well start with what Acerta knows best. The basic idea is to leverage the data generated before, during, and after the production process to derive insights into product quality or predictions about future product failures. This is most definitely a job for AI, as the sheer volume of manufacturing data being generated makes it impossible for puny human minds to grasp all the various and sundry relationships between signals.
Our clients have used predictive analytics to identify faulty transmissions, predict gearbox failures and detect anomalies in engine misfires. All of these cases involve models based on machine learning — a subset of artificial intelligence — and in each one, the models were able to deliver highly accurate results even with minimal training data. This capacity for generalization is a hallmark of AI.
2) Predictive Maintenance
Although predictive analytics and predictive maintenance are often lumped into the same category, there are important differences between them. The premise of predictive maintenance is to use data from the production line to anticipate when manufacturing equipment is likely to fail, and then intervene to repair or replace the equipment before that happens. Although it’s not a perfect analogy, one could think of the relationship between predictive maintenance and predictive analytics as akin to the one between quality assurance and quality control: the former focuses on process, the latter on product.
Nevertheless, as with predictive analytics, predictive maintenance depends on being able to synthesize insights from massive data sets, often with minimal training data. Examples of predictive maintenance using AI include machine tool builders forecasting machine spindle issues before they happen, and General Motors using image classification to identify robotic arm failures.
3) Industrial Robotics
Robots and AI go together like apple pie and ice cream, peanut butter and chocolate or Wookies and Ewoks: good on their own, but amazing in combination. Although they’ve already been in use for more than half a century, industrial robots have been changing their image in recent decades, from coldly competing against human workers, supplanting them with ruthless efficiency; to friendly helpers who can make line workers’ lives easier rather than stealing their livelihoods. At the center of this shift are collaborative robots, or cobots, which are designed specifically to work with humans.
Adding AI to cobots enables them to be deployed faster, monitor their workspaces for changing conditions and adapt to them. Regarding industrial robots more generally, artificial intelligence can improve robot accuracy and reliability as well as enable more advanced forms of mobility. Perhaps most significantly of all, AI can play a key role in reducing the programming and engineering effort required to create and implement industrial automation.
4) Computer Vision
Closely tied to industrial robotics, computer vision applications for AI in the industrial space most often involve visual inspections. Artificial intelligence has two obvious advantages over humans when it comes to visual inspection: speed and accuracy. A computer vision system using cameras that are more sensitive than the naked eye and augmented with AI can identify microscopic defects that human inspectors might miss at a rate they cannot hope to match.
To take one example, Audi used an AI vision system to identify cracks in the sheep metal from its press shop. Because this solution was based on deep learning — a subtype of machine learning often applied to large, unstructured data sets, such as images — Audi’s engineers spent months training their artificial neural network using several million test images. That initial effort paid for itself, however, since the system was able to learn independently from the examples and can now detect cracks in entirely novel images.
5) Inventory Management
Last but certainly not least, inventory management may not be the most exciting application for AI in manufacturing, but it is a valuable one. According to at least one estimate, inventory amounts to $1.1 trillion in capital. That’s an enormous amount of value that could be unlocked with better inventory management, and artificial intelligence is the key to that. There are myriad ways that AI can reduce the costs of maintaining inventory, from optimizing what’s kept on-hand to anticipating gaps before they happen.
Once again, it’s the ability to take in staggering amounts of data and find the patterns hidden within that makes AI such a natural fit for this application. Although it’s not a manufacturer, Amazon is perhaps the largest and best-known example of applying AI to inventory management.
More AI Applications in Manufacturing
There are numerous other examples of AI in manufacturing, including sustainability, waste reduction and supply chain management. The number of manufacturing applications for AI will no doubt continue to increase as computational resources become less costly and domain knowledge proliferates. In any case, one thing is certain: it’s an exciting time to be working at the intersection of AI and automotive manufacturing.
For more information about Acerta, visit our website.
