Feature: Google Cloud, GFT turn AI into production powerhouse

Employing AI and machine learning (ML) in the manufacturing process is far from a new concept, but experts from Google Cloud and digital transformation specialist GFT believe there is more to come by enabling the technologies to perceive and understand environments in real-time.The companies have developed robotic arms which can identify faulty goods, realign items on the production line and check for flaws using computer vision and sensors to mimic the actions of human workers.Fabien Duboeuf, industry lead for manufacturing at Google Cloud, and Brandon Speweik, head of industry at GFT, spoke with Mobile World Live about their work so far.

It starts at the beginningDuboeuf explained the origins of any digital transformation in a manufacturing environment involves clearly defining the problems to address and what success would look like.“Once those are identified, you should develop a robust data strategy.”“This includes determining the relevant data sources”, for example from sensors, machine logs, vision systems, manufacturing execution systems and resource planning, “collecting and ingesting the data, and preparing to clean and normalise” it so it will fit with AI and ML models.



Duboeuf (pictured, left) explained companies should not underestimate the importance of human staff in any transformation: he said they should be “involved throughout the process” along with considering what fresh skills may be required by those working with the AI set-up.“Throughout the entire process, it is imperative to implement responsible AI practices and adhere to data privacy regulations and security measures.”Speweik argues simplicity is key: “the best place to begin AI rollout is with one high value use case”, he said, pointing to visual inspections in the manufacturing process as a “strong entry point” because it employs “net-new data through images and does not require disrupting existing systems or processes”.

“From there, organisations can aggregate data across machines, sensors and operations into a common platform that can support the organisation as it scales AI rollout.”It takes brainsThere are likenesses between the system GFT and Google Cloud developed and people, with sensors feeding information to an AI brain to make decisions.Duboeuf said sensors and physical devices are the “sensory organs and limbs that give AI systems a presence in the physical world”.

“Sensors are the AI’s eyes and ears, constantly feeding it data about its environment”.“This information is vital for the AI to understand what’s happening around it.”The AI then controls physical elements including robotic arms, motors and actuators, the “muscles and hands”, he said.

“Together, this hardware combination allows an AI not just to think, but to perceive and physically respond to its environment.”Speweik added AI models for assessing quality standards are typically run at the edge of the network and operate in real time.“Once trained, the AI can immediately assess whether a part meets quality standards.

When this technology is paired with robotic arms, it can then physically remove defective components, turning digital intelligence into tangible action on the factory floor.”Duboeuf said the set-up opens the door to “intelligent automation” and the potential to address any pay-off concerns in the form of enhanced productivity, and “new levels of flexibility and responsiveness in manufacturing and other industries”.“They empower technology to act intelligently in the real world, not just process information.

”There are also clear benefits in terms of site safety, in turn advancing opportunities to use robots to augment human capabilities. Speweik (pictured, left) said human feedback also contributes to the AI’s learning, enabling the technology to adapt and further improve the efficiency of automation.It helps peopleThe GFT expert noted AI would have a dual benefit on the manufacturing sector, automating production processes along with informing real workers on how to use the system.

“When it comes to the people who will ultimately interact with and use smart infrastructure, AI can upskill employees by embedding training and subject matter expertise into the tools they use daily.”Speweik pointed to the potential for AI to “turn complex manuals into interactive learning modules” as an example.Duboeuf emphasised the potential for machinery to make decisions in advance by employing the ability of AI to analyse vast amounts of data.

“In essence, AI provides the intelligence layer that transforms traditional infrastructure into smart systems”.It wants informationSpeweik highlighted purpose-built cloud platforms as important to any AI deployment. These enable manufacturers to “bring together their data over time and power new use cases, including root cause analysis, predictive maintenance and eventually enterprise-wide intelligence”.

Duboeuf said the ability of AI to analyse vast production datasets offers the potential to “uncover subtle patterns and identify the root causes of quality problems”.Along with enabling manufacturers to improve the production process, such a set-up can conduct live monitoring, making “dynamic adjustments” to maintain high quality levels.“Collectively, these AI applications lead to improved traceability, easier compliance adherence, reduced overall costs and foster more effective human-AI collaboration”.

The bottom line is AI can turn quality management “from a reactive” process to a proactive system.Speweik adds the technology is “transforming quality control from a simple check-the-box exercise”.Duboeuf noted the benefits go beyond quality control, with AI also contributing to a broader optimisation of factories by analysing “demand forecasts, resource availability and production constraints to create optimal production schedules”.

“This minimises bottlenecks, reduces idle time” and contributes to a more efficient use of resources, “leading to higher throughput”.Predictive maintenance is another benefit the technology can enable, helping companies minimise downtime from breakdowns.It needs someoneDuboeuf and Speweik each emphasised people would remain a key element in the factory of the future, pointing to the depth of knowledge skilled staff possess and their ability to target the use of AI to areas where it is most needed.

“While AI is excellent at analysing data, human insight is often required for data cleaning, feature engineering and contextualisation needed in the manufacturing process,” Duboeuf explained.People are also necessary to validate AI model outputs and evaluate or interpret its decisions, the latter of which Duboeuf said requires an understanding of the manufacturing process.The executive acknowledged people can sometimes be the weak point in any digital transformation, pointing to a general resistance to change, gaps in skills or an unwillingness to cooperate with other teams as barriers.

He advises companies to invest in training programmes covering AI concepts and the potential the technology offers, noting some businesses have created centres of excellence to provide swift and easy access to expert guidance.The post Feature: Google Cloud, GFT turn AI into production powerhouse appeared first on Mobile World Live..