comment on this article

Staying agile

Jos Martin looks at some of the key trends expected to impact Industry 4.0 in 2020

With Industry 4.0 well under way, 2020 is expected to be a pivotal year, with engineers, designers and other professionals benefitting from the latest developments in technology, from artificial intelligence to robotics. But as the tech improves, and as more possibilities for what employees can achieve with it arise, they will also be expected to do more.

The pace of change needed to succeed in the 2020s will not slow down any time soon, so it’s vital to stay agile and constantly be on the lookout for ways to both drive efficiency and improve quality. Here are the top trends for the next 12 months that electronics designers need to be aware of in order to stay ahead of the curve.

Model-based design will be crucial

AI-driven systems are on the rise driven by new AIs being trained to work with a broader variety of types of sensor including Radar, Lidar, inertial measurement units for a wider variety of systems including aircraft engines, industrial plants, autonomous vehicles and wind turbines. The behaviour of an AI model has a substantial impact on the overall system performance, so electronics designers need to be aware that developing one is no mean feat and is only set to get more difficult and designs become more complex.

Technology can make the process of creating AI models much easier though. Model-based design represents an end-to-end workflow that reduces the complexity of designing AI-driven systems. Designers will look to model-based design tools so that they can simulate, integrate and continuously test these AI-driven systems, ensuring they are accurate and successful. Being able to simulate is important for understanding how the AI interacts in a system, and integration is critical to allowing designers to speedily identify weaknesses in the AI training datasets or design flaws in other components.

Simulation makes it easier to adopt AI

According to recent findings from Gartner, 14% of global CIOs have already deployed AI, but almost half (48%) are planning to deploy it by 2020. There are, however, a number of challenges electronics designers face before implementing it successfully into their work. One of the biggest challenges, according to Gartner, is lack of quality data, as if there is not enough data to work with, the AI is likely to fail.

Fortunately, software can be used to simulate data and supplement existing data from normal system operation so that AI models can be trained accurately, increasing the likelihood of a successful project outcome. In particular, data from anomalies or critical failure conditions is key, something especially true for predictive maintenance applications, such as accurately predicting remaining useful life for a pump on an industrial site. Being able to simulate failure data is extremely useful as creating data from physical equipment would be expensive, inefficient and destructive. Software can generate data from simulations representing system failure behaviour.

Standardised industry protocols will emerge

Industry-wide standards like OPC UA TSN will come to fruition, helping electronics designers, engineers and manufacturers ensure interconnectivity so that equipment from different vendors interoperates seamlessly.

We will also see the disappearance of inflexible cables and the influx of 5G and other wireless protocols. Standardisation will become more and more important throughout 2020 and beyond, with the continuing growth of IoT, as machines are not only connected with each other but also to cloud systems where there is arguably near limitless calculation power available.

Embedded devices become easier to deploy

Typically, artificial intelligence has used 32-bit floating-point mathematics as available in high performance computing systems like data centres, clusters, GPUs and others. This meant designers could achieve more accurate results and easier training of AI models, but that low power, low cost embedded devices that used fixed-point mathematics – such as electronic control units (ECUs) in vehicles and other embedded industrial applications – were ruled out. However, thanks to recent advances in software, there are plenty of tools available for 2020 that support AI inference models with different levels of fixed-point mathematics. The key benefit here is that organisations can deploy AI on these lower power devices and incorporate the technology into their designs.

Cloud and edge computing drive calculation power

From the workshop to the lab to the factory floor, AI-based algorithms will dynamically optimise the throughput of the entire production line while minimising the consumption of energy and other resources. Cloud computing, along with edge computing – whereby organisations process data on the edge of their network where data is generated as opposed to in a central server or central data-processing centre – is playing a key role. The rapidly increasing calculation power of industrial controllers and edge computing devices are truly paving the ground for new software functionality on production systems.

Predictive maintenance is also set to improve as it incorporates data not only from one machine or site but also from multiple locations and across equipment from a variety of vendors. The algorithms can be deployed on non-real time platforms as well as on real-time systems like programmable logic controllers.

Electronics designers can achieve more

One of the most profound changes in the electronics industry in 2020 will be on the people behind the components and machinery. As AI becomes an increasingly important part of the agenda, demand for AI skills will grow, and it won’t just be data scientists that will be sought after. Engineers, scientists and designers will need to be proactive in training themselves – or putting in a case to their employers to be trained and become ‘citizen data scientists’ – non-data-science professionals that have the knowledge and ability to work on machine and deep learning projects. Artificial intelligence is offering plenty of opportunities for improvement, but it is vital staff have the know-how to be able to fully capitalise on it.

Reinforcement learning starts being taken seriously

Reinforcement learning (RL) is possibly most famous for its ability to beat human beings at the game Go – which recently hit headlines after it beat South Korean Go champion Lee Se-dol and led to him announcing retirement, concluding that it is now “an entity that cannot be defeated”. However, in 2020 it will go from gaming to enabling real-world industrial applications, particularly control design, robotics, autonomous systems and more. RL will increasingly be used as a component to improve larger systems, with key enablers including easier tools for electronics designers to build and train RL policies, generate large simulation datasets for training of models, code generation for embedded hardware and easy integration of RL agents into system simulation tools.

One example of how RL can be used is to improve driver performance in an autonomous driving system. By adding an RL agent to improve and optimise performance, AI can enhance the controller in the system leading to advancements such as reduced fuel consumption, faster speed or quicker brake response times. RL can theoretically be implemented into any part a full autonomous driving system, from a vehicle dynamics model to an environment model to image processing algorithms and camera sensor models.

While developments in technology are putting pressure on organisations to adapt and invest in training and skilled professionals, electronics designers should look on the bright side, considering all the positives that will arise as the tech advances further in 2020.

If professionals work together to capitalise on these opportunities, they will be able to realise projects faster, more efficiently and with a higher success rate. It is already an exciting time to be an electronics designer but by the end of 2020 that statement will ring even more true.

Author details: Jos Martin is Senior Engineering Manager at MathWorks

Author
Jos Martin

Related Downloads
223377/P28-29.pdf

Comment on this article


Companies

Math Works

This material is protected by MA Business copyright See Terms and Conditions. One-off usage is permitted but bulk copying is not. For multiple copies contact the sales team.

What you think about this article:


Add your comments

Name
 
Email
 
Comments
 

Your comments/feedback may be edited prior to publishing. Not all entries will be published.
Please view our Terms and Conditions before leaving a comment.

Related Articles