Real-time data and AI thrust manufacturing into the future

The impact of real-time data-based automation on industries is becoming increasingly apparent. None are due for greater disruption than manufacturing businesses.  Sometimes harsh, these manufacturing environments in recent years have seen dramatic sensor price drops, and that in turn has opened the gates to real-time data primed for analytics. With a growing need for real-time situational awareness and insights, artificial intelligence architectures are becoming increasingly important to make sense of the acquired information.

Now, manufacturers of all sizes collect computer vision-based data at every stage of the production process. Vision sensing may be the leader in the push forward. However, pressure, heat, location and other sensors of many kinds create data streams that are digitized and stored in vast quantities. As a result, manual analysis is no longer feasible for single lines, much less entire factories or networks of factories. 

In short, the manufacturing sector is undergoing steady evolution, and the production line is going online with dramatic effect on processes, operations and efficiencies. 

The breadth of the change is sometimes overlooked, but the progress is clear. The global IoT in the manufacturing market was valued at $205.8 billion in 2021, according to Precedence Research. It is projected to reach around USD $1.52 trillion by 2030, growing at 24.91% [CAGR] between 2022 to 2030.

AI in manufacturing and related supply-chain systems is also a game changer.  According to Gartner, more than 75% of commercial supply chain management application vendors will offer embedded advanced analytics (AA), AI, and/or data science by 2026. Such systems bring AI decisions directly into complicated workflows. 

The importance of real-time 

In the manufacturing world, informed decision-making has long been vital to maintaining quality, meeting deadlines, and preventing unplanned outages, defects, or safety issues. Approaches changed in significant ways a few years ago when factories and related supply-chain systems began to tap into the Internet of Things, and to transform analog to digital processes. An industry-wide effort is underway to support the immediate response and action essential to identify and resolve problems before they escalate. 

The use of AI-based solutions in manufacturing, supply chains and logistics is ushering in a new age described as Industry 4.0, or IIoT, for the Industrial Internet of Things. The goal is to maintain the entire supply chain without any manual participation. In addition, intelligent factories powered by AI, production can run more efficiently, reduce downtime, and enhance the overall customer experience. 

For example, AI solutions like intelligent document processing (IDP) are gaining steam, helping manufacturers minimize time spent processing documents by turning unstructured and semi-structured information into usable data in real-time. Not only does this revolutionize the data capture process entirely, but it eliminates the common paperwork bottleneck that manufacturing companies see daily.

Rolls-Royce rolls in real-time

Innovative industrial companies are now processing the data they gather, often employing advanced analytical systems that not too long ago were the sole province of hyper-scale cloud providers and social media megacompanies. 

As described by Shiv Trisal, a global manufacturing industry leader at data analytics provider Databricks, AI, and data analytics are foundational to delivering more personalized customer outcomes, proactive field service delivery and differentiated mission-critical applications to their customers. An example is Rolls-Royce.

“We are collaborating with Rolls-Royce to analyze hundreds of data points per second to minimize downtime and emissions from their aviation engines flown by airlines worldwide. Manufacturers can now leverage this kind of data to operate a tech-enabled services business that demonstrates greater scalability,” Trisal told VentureBeat. 

In the past, predictions that machines were creating a defect could be made, but usually, the prediction came too late in the manufacturing process. By the time the signs of imperfections were detected, the damage was already significant enough to require the shutdown of a costly machine. 

“As per the American Society of Quality, the cost of poor quality can amount to as high as 20% of sales. Proactive detection of non-conforming materials in the manufacturing process can significantly reduce expensive recalls, lower waste, increase product quality and improve product traceability,” said Trishal. 

According to Trisal, advancements in data collection and analytics have revolutionized this process. Especially,  AI has also become a vital tool in quality control.

Using computer vision, AI algorithms can detect even the slightest defects in the manufacturing process, such as misaligned components or damaged parts. This has helped manufacturers to produce products of consistently high quality, reducing the risk of costly product recalls and improving brand reputation. 

“We have seen more and more companies using data analytics tools and platforms to successfully apply computer vision capabilities in their manufacturing plants and automate the process for quality checks, analyzing hi-resolution images at very low latency,” added Trisal. As each product moves through the manufacturing process, it delivers insights on the edge in real time to operators.

Ratcheting up ESG and workplace safety 

Another key benefit of real-time data and AI in manufacturing is the ability to improve supply chain management, including formerly paper-based processes. With real-time data, manufacturers can monitor inventory levels, track deliveries, and forecast demand, allowing them to make smarter decisions about when and how much to produce. This has reduced the risk of stockouts and overproduction, leading to lower costs and increased customer satisfaction.

“Real-time data and AI are helping manufacturing through failure prediction and maintenance planning as well as accurately identifying, contextualizing, and processing the growing volume of invoices and documents to speed along the production process,” Petr Baudis, CTO and chief AI architect at intelligent document processing platform Rossum, told VentureBeat. 

Baudis explained that from inventory management to purchasing and shipping, documentation is a true communication line between vendors, companies, and customers, and data-driven AI is the foundation that understands – and can act on – each unique format and data point.

Likewise, Kjell Carlsson, head of data science strategy & evangelism at Domino Data Lab, believes that by leveraging deep learning, manufacturers have dramatically expanded the range of inspection tasks beyond what could be automated with traditional non-machine learning-based inspection methods.

“A particularly fascinating use case here is the ability of manufacturers to mine vast quantities of unstructured data to get advanced notice of potential supply chain disruptions,” said Carlsson. “Lockheed, for example, leverages deep learning-based NLU (natural language understanding) methods to mine a vast array of data sources — for example, news information on supplier acquisitions or large orders; for early signs of potential disruptions.”   

The use of real-time data and AI in manufacturing has also increased safety in the workplace. AI algorithms fed by vision sensor systems can detect hazardous situations, such as machinery malfunctions or human error, and alert workers to take appropriate action. Additionally, using AI-powered robots has reduced the need for human workers to perform physically demanding tasks, reducing the risk of injury.

“AI helps maintain workplace safety by identifying data anomalies in real-time. Through real-time data and AI, manufacturers have the ability to consistently monitor and troubleshoot live, solving any production issues before a disruption or hazard occurs,” said Baudis. 

Moreover, the integration of real-time data and AI has helped manufacturers to reduce their carbon footprint. By analyzing real-time data, AI algorithms can detect ways to optimize energy consumption and reduce waste. This has allowed manufacturers to adopt more sustainable practices, reducing their environmental impact and helping to build a greener future.

Clearly, manufacturers at the forefront of data, analytics and AI are setting science-based targets and driving favorable sustainability outcomes today by deriving better insights from their operation, supply chain and the outcomes that their products generate for their end customers. 

Real-time data and AI form new normal 

Mike Babiak, director of supply chain tech strategy at consulting and technology company Longbow Advantage, said we will soon be seeing real-time data and AI in the manufacturing industry become the standard as opposed to being something “nice to have.” 

“Through real-time visibility (RTV), shifts, days and weeks are more successful. Warehouse managers will no longer start at a deficit. AI also helps adjust on the fly without having to depend on gut or visual cues,” Babiak told VentureBeat. 

Babiak predicts that the new normal will utilize prescriptive data and analytics throughout operations. It will also be expected that the data can work together across multiple technologies and locations and still be surfaced in real-time.

“The new development here will be the fast-growing adoption of deep learning-based computer vision models on production lines for automated defect detection,” said Domino Data Lab’s Kjell Carlsson. 

Manufacturing, for very understandable reasons, is a very conservative field, he noted. But opportunities to embed AI into processes when a line is being established or completely redesigned. 

He anticipates steady growth. “This means adoption is taking time – but it is just a matter of time,” he said.

Likewise, Rossum’s Petr Baudis believes that pursuing practical data-driven AI technology versus aspirational is crucial, especially during an economic downturn.

“For some companies, deployment can take months. If you can’t put your new robots to work in your business within the first 30 days, proving their value, impact, and return on investment can feel daunting,” he said. 

Manufacturers need AI technology that solves practical business headaches from one easy-to-use platform and requires minimal implementation time, he said. “But that’s the future.”