Industrial Automation Magazine India

5G is very interesting and could completely transform automation

Published on : Tuesday 08-02-2022

Mihir Punjabi, Director – AI/EI Solutions Architect.

Which forthcoming advances in technology will impact industrial automation? 
There are many technology advances which could possibly impact industrial automation. Few of the technologies I would like to mention here are –  
Digital twin: This is the most exciting technology in this sector. Consider it to be the metaverse of the machine world! Though it does touch upon a few hardware aspects, automation is predominantly software centric. Hence, once we have software/digital avatars of the machines/line/plant, automation will become predominant. 
 
Advanced robotics: Many organisations have already adopted robotics but are experimenting with autonomous or intelligent collaborative robotics. The goal here is to make robots as dynamic and adaptive as possible by leveraging technologies like AI/ML. For example, if a machine predicts that line is going to get halted due to a fault, then can it collaborate with other machines to start auto-offloading with minimal impact and get the line serviced all without any manual intervention? 
 
Embedded/edge AI: Due to the real-time response requirements, making decisions as close to the line will help realise many use cases. It will also enable a distributed environment of embedded devices with decentralised and quick decision making. This is one of the building blocks for advanced robotics. 
 
5G: 5G is very interesting and could completely transform the automation possibilities by enabling near edge real-time use cases which were not possible to implement earlier due to latency issues. This will be very critical to ‘intellify’ brownfield deployments. For instance, a server in the factory could handle vision-based quality inspection and recommend actions in real-time over 5G. Currently, this is being validated in lab environments by various organisations. 
 
While all these could impact the industry, I believe digital twin and edge AI would be the two interesting ones to watch out for.  
 
How are manufacturing industries leveraging AI/ML? How do automation controllers provide the necessary platform? 
I would say organisations are at different maturity levels with respect to AI/ML. 
 
Many organisations are experimenting to understand if AI can solve challenges or optimise things. For instance, can AI be leveraged to predict the quality of product before it is being manufactured?  If the predicted quality is bad, what can be done to convert it to a good product thereby reducing wastage? Another example is if AI can be leveraged to optimise efficiencies post automation – optimising HVAC controls to reduce carbon footprint for environmental sustainability. 
 
Then there are those few early adopters who have deployed AI/ML solutions like vision based quality inspection, predictive maintenance. 
 
Thus, AI/ML can be leveraged in manufacturing across various aspects from monitoring process/product to predictive and prescriptive solutions with the final goal of an autonomous factory. 
 
A major challenge here is the AI/ML accuracy and black box outcomes. A rule-based engine can guarantee 100% accuracy and output predictability which is not the case with AI/ML. Hence, industry relevant explainable AI is something that needs to be watched closely.
 
Automation controllers are the heart of all these activities. They are the ones interacting with the sensors and actual equipment. They have the two most important responsibilities – collecting the data from these equipment, and collaborator or orchestrator of all equipment based on inputs it receives programmatically. 
 
Without the above two aspects, automation would just be impossible. Automated decisions cannot be made without data and decisions cannot be executed automatically without a programmatic and collaborative interface to the equipment.  
 
AI and ML need massive amounts of data to be gathered by IoT devices. What strategies do industry plan to collaborate in data collection? 
This is the most important step since it lays the foundation of the entire exercise. If the data is not right or sufficient, the entire process goes for a toss. 
 
To enable data collection from IoT devices, factory IT and OT teams need to align and collaborate. Strategies at various levels need to be charted out – sensor installation to collect relevant data without disrupting floor operations, data pipeline and access, data security and so on. 
 
Another important factor is whether organisations are ready to publish and share non-confidential data and learnings that could help the industry overall. In an ideal world, this could be done as part of a consortium.   
 
How can AI and ML help companies create predictive models, analyse operations, make accurate forecasts and automate supply chains? 
AI/ML can learn from huge amounts of historical data and can adapt to dynamic situations. Benefits from historical data can be leveraged if operations are repetitive in nature. As part of Industry 4.0, industries are focusing on making operations repeatable, optimised and autonomous. This is enabling AI/ML models to predict and forecast accurately and bring significant value. AI/ML is already being leveraged for use cases like predictive quality, predictive maintenance, smart scheduling, and forecasting sensor values. 
 
This enables organisations to make decisions based on insights rather than intuitions.  
 
The full potential of AI and ML is realised only when the scale of operations is big enough. How can the average SME benefit with their limited resources?  
Automation is preferred only when it provides significant benefits over doing things manually. The full potential of any solution or technology is defined by the value it provides and the RoI. It is true that for some use cases, a big scale of operations is preferred but for many use cases this may not be true. 
 
SMEs should chart out use cases where they are struggling in terms of labour, time or even costs i.e., use cases with significant business benefits. For instance, if automating a production plan optimisation saves a few hours of manual effort, this could result in significant savings even at small scale. Especially, when some solution is readily available on the web, and it needs to be customised for specific needs. 
 
In addition to the RoI, the effort involved in developing the solution should also be considered. Else, this might offset the benefits entirely especially in small scale operations. Thus, readily available solutions (open source/cloud/3rd party) should always be explored. 

Relevant data availability is another important factor that needs to be considered. Approaches like reinforcement learning, synthetic data generation should be explored. Also, SMEs should collaborate with each other and share non-confidential data, insights, use cases, available solutions, maybe via a consortium. 
 
Thus, it is very much possible for SME to benefit from AI/ML with limited resources provided they can identify good business use cases and a comprehensive technical execution.   
 
The human element remains critical in deployment of new technologies. How is skill development to be planned in a scenario of not yet mature technological advances? 
I believe AI/ML and industrial automation are already part of many university curriculums while many are currently incorporating the same. Speaking of AI/ML, an ecosystem has been around for many years – I had AI as an elective in my final year engineering in India. Thus, AI/ML courses have become available quickly.  
 
Considering the entire spectrum of new technologies, I see many colleges partnering with corporate and global education institutions for quick adoption. Also, I regularly come across students proactively experimenting with new technologies like blockchain, AI, metaverse and eager to get their hands dirty on real use cases. Thus, I believe our future is bright and we need to support them by formalising these in courses, providing industry internship opportunities, etc. 
 
For working executives, reskilling is a must. Organisations need to promote and support the right executives (in planned batches) to explore new technologies and encourage them for training/certification. But I believe all this should be in a pure hands-on mode. The executive needs to build a working and presentable demo/asset as part of this exercise which should position their organisation as an early adopter.   
 
(The views expressed in interviews are personal, not necessarily of the organisations represented)
 
Mihir Punjabi is a solutions architect and a techno-business strategist who is passionate about solving business problems. Mihir is an Edge Intelligence and AI evangelist who has presented at various industry forums and is currently responsible for AI solutions in Manufacturing within the Capgemini Engineering group. The views expressed are personal.