Industrial AI in action: Creating business value at scale in smart manufacturing

The combined power of OT and IT data and domain-specific AI and ML models deployable in versatile environments is enabling manufacturers to unlock new levels of efficiency, productivity, and competitiveness.

However, there are challenges to overcome on the path to realising the full potential of smart manufacturing. In this article, and based upon the conversations we had at the Manufacturing Digitalisation Summit, we will explore the strategies to overcome these challenges and drive enterprise-wide transformation.

Smart manufacturing refers to the use of cutting-edge technologies, like advanced digitalisation investments and analytics on important data, to optimise the entire manufacturing process, from production to distribution and maintenance.

These technologies enable manufacturers to make data-driven decisions, automate complex tasks, and optimise operations, resulting in lower costs, enhance quality control, and accelerated time-to-market. Increasingly, its success is resting on gaining a heavy dependence on culture and a crystal-clear focus on business outcomes.

The adoption of advanced tools, solutions and processes in smart manufacturing is gaining traction across industries. It is projected that the global smart manufacturing market will grow from $297.2bn to $787.5bn by 2030, at a CAGR of 14.9% (source: Grand View Research).

According to MarketsandMarkets, use of AI in manufacturing market is projected to reach $16.7bn, at a CAGR of 47.9% from 2022 to 2027. These figures highlight the immense business value of smart manufacturing practices.

The roundtable approach

In a roundtable discussion at the Digitalisation Summit, using a “4L” retrospective methodology of what the leaders present Love, Learn, Lack and Long (for), participants were asked to rank, rate and elaborate on their experiences and preferences in key areas in smart manufacturing, namely:

1. Product Adoption and ROI

2. Change Management

3. Smart Technology in Manufacturing, like Digital Twins

The most popular topic was, indeed, Change Management.

Achieving ROI and the complexities of product legacies

One of the key challenges in scaling smart manufacturing initiatives is the complexity of integrating diverse systems and technologies. According to Deloitte, 44% of executives identified legacy systems as a barrier to implementing advanced solutions.

Modern manufacturing environments are complex mix of disconnected systems that do not communicate with each other. This lack of interoperability hampers data sharing and insights, preventing organisations from harnessing their full potential and measuring ROI.

A Frost & Sullivan survey revealed that 79% of manufacturers face challenges in integrating legacy systems with new technologies. To overcome these, companies need to invest in a robust digital infrastructure that can seamlessly connect and integrate various systems.

Change management

Change management is a critical factor that can and has sunk many digital transformation projects in smart manufacturing. Deep discussions with multiple stakeholders in a manufacturing organisation has revealed a near unanimous conclusion: if everyone in an organisation is not brought on board with clear benefits and work expectations caused by any changes brought on by digital transformation, the initiative will fail. One example cited a small change to well-intentioned a initiative called Dynamic Scheduling.

Most scheduling is reactive, where work is carried out in response to production needs. The shop floor adjusts to complete work as it comes in, whereas Dynamic Scheduling is proactive, with the schedule being adjusted to maximise production. A dynamic scheduling system adjusts production to minimise resource issues (like machine breakdown, tool failures, quality issues) or job-related (rush jobs or cancellations) and ensures optimal use of shop floor resources.

In one example cited by a participant, a manufacturing organisation had implemented a major investment in a dynamic scheduling system. However, it failed to give the organisation the results they needed, and was abandoned after much time and cost had been sunk in the project.

The reason was that the change didn’t take into consideration a critical human element: it required employees that worked on a particular assembly line to log in via a keypad or card swipe to indicate work start. Without it, the dynamic scheduling system wouldn’t work. There were no repurcussions or any incentives explained to the workers to do so – they saw it as adding no value, especially considering they had already swiped in for the day at their workplace.

In another example that this author has witnessed, the use of an IoT sensor to detect a changes in railway car door motors’ phase current was proposed to implement predictive maintenance. Repeated phase current variation detections would provide prognostics into a motor’s health. However, even though the actual cost of adding a new sensor was less than two dollars, the change to the workers and workflows were immense.

First, it meant their methods of recording their work had to be digitalised, on a tablet. Records had to be digital, as the sensors were sending signals digitally. The change in paperwork and recording procedures meant a revision to long-standing labour union laws. This meant involving the rail operator’s Human Resources, Legal, and IT departments, which was estimated to take more than two years to implement.

Eventually, even a simple and cheap change to digitalise had far reaching corporate implications, and the initiative was abandoned after significant investment in research time and costs.

Obstacles abound, but a careful analysis of changes to all stakeholders is essential to achieving a digital transformation process that sticks.

Smart Manufacturing Technologies

1. Predictive Maintenance (PdM)

Traditionally, manufacturers use reactive maintenance, fixing equipment only when it breaks down. By leveraging AI/ML-driven algorithms, including by analysing historical data, manufacturers can predict and prevent equipment failures before they occur.

These algorithms can identify patterns and anomalies, enabling manufacturers to schedule maintenance proactively, optimise spare parts inventory, and minimise unplanned downtime. PdM also prevents waste by precluding the need for unnecessary maintenance.

2. Digital Twins

A digital twin is a virtual representation of a physical product, process, or system. By leveraging real-time data from sensors and ML algorithms, manufacturers can create a digital replica that mirrors the physical counterpart.

This digital twin can simulate and optimise various scenarios, predict performance, and improve decision-making throughout the product lifecycle. The use of the digital twins is projected to reach $183bn, at a CAGR of 41.6% during the forecast period 2023–2030 (Source: Meticulous Research).

3. Quality Control

Another key application of AI and ML in smart manufacturing is quality control. AI and ML algorithms can be trained to analyse vast amounts of data, including images, sensor readings, and production parameters, to detect defects and anomalies in real-time. By automating inspection processes, manufacturers can achieve higher accuracy and consistency while reducing the need for manual inspection.

According to a study by Capgemini, AI-based quality control systems can reduce the cost of quality by up to 50% and increase productivity by up to 25%. While PdM, Digital Twins and Quality Control are examples where plenty of good technology and tools exist, few have found repeated success and large scale.

In reality, digital industrial transformation is a very high-touch activity: it takes much more than a software product or singular service that enables long term, repeatable success. It takes a bridging of the worlds of IT, OT and business operations with a portfolio of factory-hardened industrial IoT solutions, versatile and easy to use software products, and professional system integration and advisory services and support.

Secure IT + OT Data Systems

As the use of IoT, analytics, digital twins and other advances facilitated by better gleaning of insights from contextualised IT and OT systems data proliferates, their impact on smart manufacturing will become even more profound.

However, continued success of AI and ML requires a wholesome approach, including investing in talent, data infrastructure, and robust OT cyber security. As more connected devices gather larger amounts of data, implementing robust cybersecurity measures to protect information needs to be a focus.

The risks are significant. According to IBM, sixty-one percent of cybersecurity incidents at OT-connected organisations last year were in the manufacturing industry. Implementing strong cybersecurity and establishing strong governance frameworks are essential to protect sensitive data. This includes implementing strong access controls, encrypting data, updated software and firmware, and conducting rigorous vulnerability assessments.

Additionally, organisations should establish a strong governance framework to define policies and procedures for data handling, storage, and sharing. It’s the only way manufacturers can build trust and confidence among customers, suppliers, and other stakeholders, thereby enhancing the staying power of smart manufacturing technologies.

---

About the author

Shamik Mehta is the Director of Industrial Digital Services at Hitachi Vantara with 25 years of experience in IIoT, AI/ML-based data analytics, semiconductors, renewable energy, and e-mobility. He specialises in thought leadership for technology applications in Smart Manufacturing, Energy, and Electrified Transportation.