Manufacturers need every edge to stay competitive and artificial intelligence (AI) is being touted as one of the key technologies to achieve that. Mark Venables looks at the opportunities and challenges for AI within the sector.
We live in a connected world, from phones controlling the temperatures of our homes to cars communicating with one another, and to smart web-based systems. Turn on the news and you will see just how these new technologies are being used to improve the way we live.
It is therefore not surprising that the manufacturing sector is also starting to benefit from these technological advances to improve production performance. “Today, investing in technology is likened to success,” Martin Walder, VP of industrial automation at Schneider Electric, says. “Manufacturers who choose to modernise will gain increased outputs, higher quality and less wastage on the factory floor, ultimately allowing them to improve profitability and succeed in the industry.”
Industry 4.0 and associated technology, such as the Internet of Things (IoT), AI and robotics, have become part of the manufacturing vernacular, without many understanding their potential. “Digital transformation offers unmatched potential for manufacturers,” Walder adds. “Not only does it vastly improve communication between devices, systems and personnel both inside and outside of the company, but it also provides the directness to cut energy consumption, increases efficiency, and increasingly delivers even short-term ROI.
“Artificial Intelligence is one technology that will revolutionise the field. A recent report by Accenture showed corporate profits are said to increase by an average of 38 per cent by 2035 thanks to the advanced deployment of Artificial Intelligence into financial, IT, and manufacturing applications.
“In the UK we’re in the early stages of AI implementation in manufacturing. We lack in clarity as to its deployment across multiple use cases. However, many organisations are evaluating potential risk and reward scenarios, and the technology is becoming more widespread. Investing early, as with digital transformation pays dividends, but there are some crucial lessons to follow.”
AI has the potential to exponentially increase the productivity of our industrial assets. It represents a new way for humans and machines to work together in industrial applications. However, in these scenarios, many variables need to be accounted for to achieve a successful and competitive outcome.
“On the factory floor, AI technology enables us to learn and predict tendencies to solve complex problems,” Walder continues. “For example, managing a process with almost countless variables, such as control of temperatures, pressures and liquid flows, is very prone to error. In almost all factory settings there are too many variables for any human brain to analyse successfully. By implementing AI, crucial operational decisions can be supported in real time, greatly improving safety, security, efficiency and productivity.
“The quality of the data that trains the AI algorithms needs to be combined with the human expertise, which is always needed for interpretation and guidance. For example, in the food and beverage industry, AI can improve quality inspection, providing humans with vision analysis and sound analysis which goes beyond the ability of a human alone.”
Transforming Industry 4.0
AI is becoming an important part of Industry 4.0. It brings with it the great potential for innovation to dramatically increase the productivity of industrial assets, better manage the evolution of the workforce, and greater energy efficiency.
“Let’s take discrete and process manufacturing as an example,” Walder adds. “Here, asset maintenance is one of the industrial processes that is emerging as an early AI application. As a result, we’re seeing more manufacturers understand that predictive maintenance can be blended with the more traditional approach or preventative maintenance. The two work hand-in-hand.”
A great example of how AI is revolutionising Industry 4.0 and improving efficiencies on the factory floor is Variable Speed Drives (VSDs). VSDs are connected to motors on the factory floor. They attain data and insights into abnormal behaviours and thus flag these issues so that they can be repaired, or where necessary, replaced. The benefit here is that a piece of equipment on the factory floor is only replaced when absolutely necessary, saving the manufacturer money and reducing operational downtime. Machine learning also comes into play here. It can be executed at the edge to help in the early identification of many potential faults including power generation turbine blade damage or plant motor coupling approaching failure.
Decentralised Artificial Intelligence
As AI in the marketplace becomes increasingly controlled by just a handful of big companies that own cloud-based AI platforms and APIs, the issue of trust is stimulating growing calls for the decentralisation of AI. “The main fear for manufacturers is that a centralised model will lead to the monopolisation of the AI market,” Andy Coussins, senior vice president and head of international at Epicor Software, says. “This in turn could cause unfair pricing and stifle innovation.”
Decentralised AI – born at the intersection of blockchain, on-device AI, and IoT – helps solve this challenge and promotes transparency. “It also ensures interoperability and encourages innovation among an unlimited number of other AI companies,” Coussins adds. “Ecosystems such as SingularityNET are already fostering wider collaboration among the global decentralised AI community, a case of safety in numbers, if you will. More than that, such marketplaces have been designed to ensure that, in the event of AI reaching mass market usage, contributors and users of the technology will be the ones to control it, rather than a few powerful entities.
“Promoting interoperability and decentralised AI will ultimately lead to an era of AGI (artificial general intelligence) that will empower manufacturers. For example, by helping them detect anomalies and generate predictions that can be used for enterprise resource planning (ERP) and help improve their processes in the future.”
The Search for Manufacturing Efficiency
Over the last five years the automation industry has been asking the question ‘how can production lines avoid running into long changeover-times or other inefficiencies for good production?’ Manufacturers are realising that machines can no longer be set up in a fixed, inflexible manner on the shop floor, where they are commissioned, parameterised and tuned for one specific product that is produced over and over again for months or even years.
“AI algorithms are improving the efficiency of the entire factory production line, reducing energy consumption and waste, enabling organisations to meet important corporate social responsibility targets as well as deliver invaluable cost-savings,” Jos Martin, senior engineering manager at MathWorks, explains. “However, being able to capitalise on AI for production optimisation relies on investment in the latest industrial controllers, edge computing devices and cloud systems which allow engineers to take advantage of greater calculation power and functionality of production system software in the first place.”
Tomorrow’s production lines must be flexible, built from multiple mechatronic modules that can easily be rearranged, with more and more robots or cobots, and an AI that parameterises and tunes the machines according to the next individualised good that is manufactured on the line. “While AI is improving the accuracy of predictive maintenance applications such as those for predicting the remaining useful life for an industrial site pump,” Martin continues. “A major barrier to adoption is having enough high-quality data to train AI models properly. Lots of failure data is needed to ensure the AI model is accurate, but it is expensive and inefficient creating data from real, physical equipment.
“There will be more challenges in deploying AI in manufacturing, for example there will be increased need for engineers and scientists, not just data scientists with the skills necessary to build AI systems and carry out AI projects. Fortunately, if they run into any problems they are struggling to solve on their own, they can turn to an existing body of knowledge – such as deep learning models and accessible research from the community, giving them a significant advantage to starting from scratch.”
Intelligent Supply Chain
The advancement in AI technology is delivering real benefits to manufacturers, particularly when it comes to supply chain management, productivity and cybersecurity. Currently, 22 per cent of manufacturers are already using AI and given the benefits that AI provides, the usage is likely to only grow.
“As AI becomes more popular, this will impact the type of work that people do on a day-to-day basis, by allowing organisations to free employees up from their regular mundane tasks and give them the chance to add value to the business,” Graeme Wright, chief digital officer for manufacturing and utilities at Fujitsu UK&I, says. “For example, if it floods in the rainforest, people cannot harvest trees and the price of paper significantly increases. But if employees have a fully integrated system in place, using explainable AI, it will tell them that there are large amounts of rain expected within the next seven days.
“This would then enable purchasing to lock down the paper price with suppliers, saving the business from having to pay the increased price, and most importantly enabling the business to keep consumers happy by not passing on the additional cost. This is ultimately what provides employees with the opportunity to take up roles that are more engaging and enables humans to take the right decisions, proactively plan their next steps and take calculated risks.”
Cybercrime and hacking are also enormous issues for the industry, in fact it is one of the biggest challenges organisations face with nearly 50 per cent of manufacturers being a victim of cybercrime. “As the sector becomes increasingly connected and data driven, every single manufacturer has a responsibility to make sure data, systems, and connected infrastructure is protected; and AI can play a crucial role in strengthening cybersecurity and fending off attacks,” Wright explains. “By turning data into actionable intelligence, organisations can recognise more threats and respond much more effectively and efficiently to protect their businesses. Only by doing this can manufacturing companies avoid the financial and reputational risk that comes with suffering a cyberattack.”
The Role for ERP in AI Deployment
According to IFS, half of manufacturers will be using some form of AI by 2021. But according to a study released in January by Plutoshift, manufacturers are struggling to adequately adopt AI across their business. Sixty per cent of respondents said their company has been unable to come to a consensus on a focused, practical strategy for AI implementation, while 72 per cent said it had proved harder than expected to set up the technical and data infrastructure necessary to make the project viable.
One factor contributing to the difficulty is that the ERP software running their business has not previously been capable of facilitating their AI journey. “Many of the manufacturers that have had success with AI tools have done so by solving specific problems in isolation, such as demand forecasting, supply chain optimisation, schedule optimisation or natural language processing (NLP)-driven customer service bots,” Bob De Caux, VP AI and RPA at IFS, explains.
However, ERP tools with comprehensive AI capabilities embedded within them can collate and analyse data from every facet of a manufacturing organisation, helping them to accurately plan ahead, optimise processes, and reduce waste. “Using ERP as the vehicle for AI, manufacturers can reinvent their business around digital-first processes and disrupt their market,” De Caux continues. “The AI-enabled ERP allows manufacturers to optimise or automate end-to-end processes rather than just specific tasks, streamlining a quote-to-cash value chain or sales order capture-to-shipping. By combining classic ERP datasets, such as maintenance history with streaming data from assets and external data such as economic and weather indicators, companies can forecast and optimise margins from operating diverse asset portfolios like power grids or offshore oil rigs.”
A Tool for Maintenance
No one in industry can ignore the ubiquity of the IoT. Embedding intelligent IoT technologies into machines promises to transform run-of-the-mill manufacturing facilities into smart factories. “The impact will be seen across all aspects of operations from asset management and machine maintenance to planning, quality control, and even field service,” Rafi Billurcu, partner, manufacturing at Infosys Consulting, explains. “The introduction of smart sensors, one such IoT technology, promises to enable the real-time monitoring of machines – but predictive maintenance is not possible with sensors alone.
“The key to realising the full potential of IoT is to employ artificial intelligence technology alongside it. In order to put the vast amounts of data generated by these IoT sensors to good use, organisations must invest in AI as well – or risk failing to make predictive maintenance worthwhile.”
Using AI means maintenance can be predicted based on previous machine data, mapping it out across a longer period of time and spotting anomalies and changes that an average human could not. “What’s more, integrating machine learning into a system’s analytics capabilities can increase the accuracy of the predictive algorithms, enabling a more sustainable system that can learn over time, offering an even better return on investment,” Billurcu adds. “Although there is a high upfront cost, employing these technologies will mean engineers can avoid unplanned downtime and minimise negative consequences.
“Applied across the entire supply chain, this technology could save millions. Run-to-failure won’t cut it in the smart factory of the future – it doesn’t make business sense. Predictive maintenance, backed up by AI technology holds the key to streamlined supply chains that are imperative for success in Industry 4.0.”
AI is not a new concept for the manufacturing industry, but it has certainly crossed into the mainstream in the last couple of years and is now more accessible to smaller manufacturers. Although it is still not as widely deployed as other industries, it is having proven success in a number of areas.
Research from McKinsey found that AI used to monitor and analyse factory equipment assets and machinery reduces machine downtime by up to 50 per cent thanks to the analysis of multiple data points and historical data to forecast possible service requirements. That means field service teams can deliver maintenance before any machinery breakdowns. This not only reduces downtime per machine but also extends the machine’s life span by up to 40 per cent. McKinsey estimates the global financial savings from predictive maintenance at $0.5-$0.7 trillion.
“AI also provides manufacturers with data to drive further improvements and insights,” Amit Jain, senior vice president of product for ServiceMax says. “However, the sheer unprecedented volume of data being generated today, which is set to continue to increase almost exponentially moving forwards, is simply too vast to be useful unless we implement AI within FSM systems. This is also the case with interpreting IoT data, which is largely predicted to be the backbone of field service operations of the future and is empowering service organisations to move away from the traditional break/fix approach to much more effective and profitable advanced service models.
“Solutions such as ServiceMax Remote Triage help service organisations by analysing data from multiple sources to reduce unnecessary dispatching of technicians and service equipment – what we refer to as ‘truck rolls’. Unlocking insights with natural language processing and machine learning, service organisations can empower their service teams with a guided process that asks pinpointed questions and ranks possible solutions by likelihood of success and cost effectiveness. Service stakeholders can triage the problem with greater accuracy, recommend a remote solution or, if a truck roll is required, have greater confidence in selecting the right technicians for the job and dispatching them with the right parts and tools. Targeted, purpose-built AI solutions such remote triage offers field service teams in manufacturing an easy and risk-free entry into the world of AI technology.”
While it is challenging, some manufacturers are already making progress with AI for predictive and prescriptive maintenance, resulting in less unplanned downtime, more efficient operations and better compliance with health, safety and environmental (HSE) regulations. “The key to predictive maintenance is determining which data, collected from machines in operation, can be used to predict future events,” De Caux says. “Everything from vibration to heat and power draw data and may be used as the raw material upon which AI algorithms and stochastic methods can build, predicting breakdowns, diagnosing issues and enabling advanced reliability-centred maintenance (RCM). In addition, by modelling and simulating processes through a digital twin of a production facility or piece of equipment, companies can get improved visibility over a variety of scenarios and leverage AI to receive recommendations for how best to handle them.”
Technical and non-technical hurdles
However, how people view the impact of AI remains a major barrier with a third of leaders worried how employees will deal with future change. Organisations must engage their employees, and this is where it is important to properly evidence the benefits of AI and all it has to offer. “By doing so, organisations can drive the right behaviours throughout the business,” Wright explains. “But they should not have to do so alone; picking a strategic partner that knows technology, change and benefit realisation methodology means they will be able to better unleash the data which can provide solutions across the whole business. At the end of the day, people do not fear change, they fear revolutionary, unexplained change, so implementation needs to be an evolution with the benefits clearly communicated.”
Even the most elegantly designed AI algorithms need data, and a lot of it, to learn from. Many manufacturers may struggle to supply enough to build their own models, particularly if their historical data is stored in distributed or siloed environments with different data models and conventions. “An ERP system can facilitate this process for customers through sound master data management, as well as providing models pre-trained on large data sets that can help drive value for end users right out of the box,” De Caux explains. ”Before even reaching this stage, the manufacturers must overcome non-technical hurdles, including identifying which business problems, data and algorithms are meaningful, and determining if these factors will continue to be significant in the future.
“Selling AI to the C suite can also be a challenge, as senior business leaders may not immediately understand its full potential. The focus must move past immediate cost savings to new and transformative ways of driving value that were not possible without AI. In order to do this, companies need bilingual citizens who can understand the business problems to be solved and act as the glue between the business and more technically-oriented data scientists.”
Once a strategic direction is identified, the focus must be on change management. Employees will need to understand and own their role in the AI transformation, while customers and other stakeholders must be educated and mentally prepared for what is to come. “Stakeholders may have different fears, a loss of human contact with their vendor, diminished employment opportunities or, in the case of senior management, endless investment with no firm payback,” De Caux adds. “Employees need to see that once AI takes on mundane tasks, they can be free to concentrate on higher value responsibilities, while customers need to know these newly liberated employees will be able to spend more meaningful time with them. Finally, senior management must lead the change by finding ways to measure the value that is brought by this new technology, using AI-driven ERP to update standard metrics, KPIs and ROI calculations to reflect the new business reality.”
The Benefits of AI
According to Dan Jelfs, SVP of global sales at Mobica, the potential benefits of AI and machine learning will be wide-ranging and numerous.
With AI, manufacturers will have the capabilities to make sense of huge volumes of data provided by sources, such as the Internet of Things. This technology will underpin Industry 4.0, helping companies to limit operational disruption, drive efficiencies, reduce costs, enhance performance and increase speed to market.
For example, AI will enable production managers to develop prediction failure models and carry out impact analysis which will, in turn, allow for preventative maintenance and the optimising of processes – ultimately, eliminating unnecessary downtime.
With an AI-driven approach, manufacturers can also introduce computer vision and sensor-based technology that will not only improve quality control but enable the deployment of advanced operational technology (OT), such as collaborative robotics. It will also power digital twinning which manufacturers can use to plan and test designs in a virtual environment as well as run operational simulations.
Within the next decade we will undoubtedly see AI technologies achieve mass market adoption. However, there are several challenges that need to be overcome for this to become possible. These include the convergence of various forms of IT and OT, the managing of data at scale and the need for mature digitalisation programmes. Robust security in AI enabled manufacturing systems and supply chains will need to be developed. It’s worth noting that there are still uncertainties around how AI technology will be rolled out and how it will be impacted by the emergence of edge AI capabilities.
Manufacturers then need to consider the human factors. As this is still an emerging technology, a company will need to access a pool of talent with the appropriate skillsets in this discipline in order to keep exploring its potential. The adoption of this technology will also fundamentally change operations so will need to be considered in the overall corporate strategy, championed from the top of the business and rolled out with an effective change management plan in order to drive the cultural shift required and address any employee resistance.
As AI matures however, we can expect to see its impact driving positive change in supply chains and logistics, right through the product lifecycle – from raw material to end of life recycling. It’s also likely that AI will provide manufacturers with greater accessibility to blockchain-driven solutions, as the industry strives to operate ethical supply chains in a bid to achieve global sustainability.