All News & Insights

Disruptive Technologies: Moving Manufacturing into the Future

Advanced technologies and processes such as artificial intelligence, machine learning, and digital threads can help manufacturers cope with changing customer requirements and increased global competition. By Atul Mahamuni


The manufacturing industry is facing many challenges today. From higher expectations from customers, to sustainable business practices, the manufacturing industry is dramatically transforming. Now, the adoption of new technologies is imperative for businesses to stay competitive in the age of increased global competition, elevated customer expectations, acute labor shortages, and ever-changing regulations. And, having access to real-time data and the insights it can provide can help avoid risks and prepare businesses for the future.

With so much innovation taking place in the manufacturing industry, it is important for businesses to get more insight into what transformative technologies are driving the modern manufacturing movement. In this article, I will discuss the transformative technologies and some potential best practices for selecting and deploying these technologies in manufacturing.

Drivers of Disruption 

In today’s factories, smart automation is driving significant change, leading to the smart, connected factory. What makes this smart automation possible? The answer includes three distinct technologies. First, the connectivity technologies of the Internet of Things (IoT) are making it possible to collect data from factories with much less effort and cost than before. Second, machine learning and artificial intelligence (AI) technologies are making it possible to process large amounts of data and gain insights to further predict outcomes and potential issues such as machine failures, component shortages, and quality issues.

Third, manufacturing processes are increasingly being integrated with other business applications to automate end-to-end business processes. This integration of physical and cyber systems, which offers the potential to help eliminate errors and costs associated with manual processes, is enabling the age of smart automation at a rapid pace.

Globalization levels the playing field for all manufacturers – domestic and foreign. Simultaneously, globalization also increases competition for all companies, forcing each to out-execute their competition for sustained profitability. In addition, globalization creates an ever-increasing and diverse set of compliance and regulatory requirements. Adoption of technologies is frequently the only way to stay ahead of the competition.

Consumption models of today’s consumers are changing, and the popularity of newer business models such as servitization – selling the products as-a-service – are rapidly increasing across all types of products.

Fortunately, these disruptions can be tackled using a slew of emerging technologies. It is important for businesses to understand the primary technologies that enable a best-in-class IoT solution.

Real-Time Visibility 

Functional and operational managers often describe their jobs as walking a tightrope as they precariously balance multiple responsibilities protecting their operations against a constant barrage of things that go wrong. To make matters worse, they often lack the real-time visibility that helps them get a consolidated business view of their assets and operations. This lack of knowledge is exacerbated because the information they do have is often disconnected from the physical environments they’re working to manage.

Digital twins of assets, factory floor machines, and processes help provide real-time visibility. They empower users by creating a 360-degree view of their assets – from the sensor values, to KPIs, to incidents, to maintenance or financial information related to that asset.

Digital twins also provide a digital interaction model of the physical assets that can be used to get the historical, present, and future views of the assets. This interactivity model also enables the ability to control the assets remotely.



The integration of physical and cyber systems is enabling the age of smart automation at a rapid pace.

Usability of this digital interaction model is enhanced significantly with augmented and virtual reality technologies. The user simply looks at a machine through a handheld tablet or smart-glasses, and, in an instant, the digital twin of that machine comes alive, providing real-time views, KPIs, maintenance events, financial information, and failure predictions.

The biggest challenge in dealing with digital twins is that most companies have a very narrow view of what a digital twin is. For some, it is just a CAD or a finite element analysis model. For others, it is just a virtual reality interface. While these are important aspects of a digital twin, characterizing one to this narrow scope is sub-optimal. A complete definition of digital twin includes the following facets:

Virtual Twin
The first facet is a virtual twin, defined as the software (virtual) representation of the physical asset. Companies use this facet to improve interactivity with the physical asset. For example, customers can visually see the asset performance parameters on their screens. They can drill into subcomponents of the assets and look at the real time status of various attributes of that subcomponent. In addition, they can also control the asset remotely by issuing commands simply by configuring the digital twin.

Predictive Twin
The second facet is predictive twin, defined as the behavioral and predictive model of the asset. Companies use this facet to predict various KPIs of the asset. For example, many customers use the predictive twin to understand machine failure probabilities and take proactive corrective actions. Other customers use the behavioral model to detect anomalous behavior patterns automatically. Customers can also use this to ask what-if questions to understand how their systems would react in the event of an unusual situation.

Twin Projections
The third facet is twin projections, defined as the integration of the insights derived from the above two facets into business processes. Companies use this facet to automate responses to various reactive, as well as predictive, events in order to extend the reach of their business systems to the physical assets.

Analytics, Machine Learning, and AI 

Smart factories outperform and out-produce today’s traditional factories through advanced machine learning power and artificial intelligence algorithms which add the necessary brainpower to make manufacturing intelligent. Advanced analytics most significantly impact the following areas of manufacturing:

Automated Business KPIs
KPIs that are used for driving day-to-day business decisions are calculated automatically so that decision makers have an instantaneous and historical perspective of the KPIs. Further, comparative analytics capabilities from AI and machine learning enable factory managers to pinpoint which factories, machines, or products are performing sub-optimally relative to others. Having this knowledge ready at the fingertips improves the quality and speed of decision making.

Anomaly Detection
Today, most instances of deviations from normal behaviors go undetected for two reasons: lack of constant manpower required to monitor every aspect of production, and the general inability of the human brain to detect all patterns of anomalies based on multiple points of data. With advanced algorithms, we can automate the continuous detection of anomalies on thousands of data points. Every instance of an anomaly can then be analyzed to decide whether it is actionable or not.

This is where machine learning and AI get the most attention. If we could predict machine and process failures in manufacturing, then we can virtually eliminate unplanned downtime. The ultimate goal of continuous high-quality and high-yield production is achievable through the smart use of predictive algorithms.

While predictive algorithms can predict probabilities of failures, the prescriptive algorithms recommend the best course of action in order to deal with a probable situation. These insights eliminate the guesswork from the operations and often provide optimal solutions.

Last, and probably the most influential, is the ability to forecast the entire supply chain process based on multi-dimensional trends. This is where the entire process of manufacturing becomes a learning system that learns from the results from every shift of operation, every run of production, every batch of raw-material consumed, every control input applied, and every sensor input. This continuous learning system makes increasingly smarter forecasting decisions to take manufacturing to its fullest potential.

The Role of the Digital Thread  

Thinking of IoT in isolation is the primary reason why a large percentage of IoT projects fail. Instead, it’s critical to think about IoT as a way to extend business processes to the physical world. Every IoT deployment in the manufacturing plant must start with a deep functional understanding of manufacturing processes.

To truly be competitive, today’s manufacturers must break down internal silos and barriers that exist between information technology (IT) applications and operational technology (OT) and extend business applications, such as manufacturing or planning operations, to physical systems through a process called IoTification. IoTifying business applications can improve new product designs based on the data collected and on how today’s products are manufactured, transported, installed, used, and serviced. It means using the information from existing operations to create a new generation of products, or even to launch a new transformative offering based on consumption economy.

As a result, IoT technologies create a common digital thread that runs through all operations of the supply chain – product design, sales and order management, procurement, manufacturing planning and execution, quality testing, transportation, warehousing, logistics, and monitoring of the product conditions as it is used by the customer.

Creation of the digital thread through integrations between various IT and OT systems works in two ways. First, it automates processes across various functions. For example, if the predictive algorithm provides a warning that a critical machine is likely to develop a fault, we could trigger reallocation of machines to the production lines or even change the manufacturing plan. This automation across multiple functions through a unifying digital thread significantly impacts manufacturing. Second, as more processes are interconnected, the digital thread becomes a mechanism to collect more data across disparate systems. Smart automated algorithms can then find any correlations between an event in one system and its impact in another system. This allows the systems to become self-optimizing over a period of time.

Best Practices for Adopting New Technologies  

While these technologies have tremendous potential to make modern manufacturing efficient and optimal, it is not often easy to adopt these technologies. There are many pitfalls in the journey, and following best practices maximizes the probability of success in manufacturing transformations.

The best way to ensure success in a manufacturing transformation project is to start with clearly defined business outcomes. Quite often, the lack of clear definition of business outcomes makes it difficult to justify the project, or sustain it, when inevitable roadblocks are encountered in the transformational journey.

What does a business outcome look like? Well, you could decide that you want to decrease the unplanned downtime by 15 percent. Or, you could decide that you want to improve the Overall Equipment Efficiency (OEE) by a certain percentage. Other examples involve improving quality or yield of production.

Once that business outcome is defined and agreed on, let that goal drive further discussions in terms of what machine learning/artificial intelligence algorithms you need. For example, if your goal is to reduce unplanned downtime, then you may decide that you need predictions of machine failures and recommendations on optimal machine reassignments. Instead, if your goal is to increase OEE, you would focus on the instantaneous as well as predictive computations of OEE and on finding the correlation between various contributing factors that affect OEE.

Once you have a clear vision of the smart insights that you need from your machine learning algorithms, that will also define what data needs to be collected and from which systems.

Staying focused on the business outcomes makes the project flow in logical steps with the ability to justify every subproject by showing how it is related to the primary business goal.

When selecting vendors, look for those who have the functional knowledge of the business applications and the thought maturity of starting the transformative project based on the business outcomes.



“The biggest challenge in dealing with digital twins is that most companies have a very narrow view of what a digital twin is.”


Don’t Underestimate Integrations 

The smart algorithms work well only when they work on clean real-time and historical data. Even if data is available in other business applications in your organizations, it takes significant effort to get the data in the right form using the right integrations at the right frequency, and connect it correctly with the smart algorithms.

It is very easy to underestimate the effort required in interconnecting these systems and making the right data available at the right time. Many projects don’t yield accurate predictions or other smart insights due to inaccuracies in the data. Other projects suffer an ever-increasing scope-creep to fix one data stream after another.

Gartner has predicted that by 2021, three out of five factory-level AI initiatives in large global companies will stall due to inadequate skill sets.

Data science models can take a long time to be built and people trained on them so that they work in your environment. And these models would continue to serve us well as long as the environment remains exactly the same (which almost never happens). A lot of things change in this era of changing customer demands and consumption patterns.

The only right way to solve the problem is to automate the process of building the data science models. While there are many software vendors who claim to have a large menu of machine learning models, this do-it-yourself approach requires sophisticated data science skills and significant time leading to failures of many projects.

Picking the Right Vendor 

To ensure success, you need to be working with vendors that have created smart software that automatically selects an appropriate data science model with the right set of inputs, tunes its parameters, deploys it in live operations, and then continuously monitors its performance and accuracy. Of course, sophisticated software is required to make this all work autonomously, but taking a path of trying to recruit and train data scientists will distract you from the primary mission of your organization.

There are two ways to embark on this journey: The hard way or the easy way.

You can take the hard way and adopt a bunch of enabling technologies. You may even be led to believe that some of those are best-of-breed technologies. A recent industry survey found that over 60 percent of transformative projects fail. A primary reason for those failures is that it is very difficult, time-consuming, and expensive to actually assemble all these technologies, along with your existing business applications to drive business outcomes. Even if you succeed in getting there, there are further recurring expenses in maintaining that system over a period of time.

The easy way includes selecting pre-assembled products to solve your business problems. These products are often consumed as software-as-a-service, virtually eliminating all the costs and pain associated with building a custom solution from the ground up.

We have discussed the business drivers that are forcing the transformation in modern manufacturing. As previously stated the adoption of various technologies such as digital twin, smart analytics, and digital thread all help in the journey towards modern manufacturing. The best practices to tackle challenges include: focusing on the business outcomes, not underestimating the integrations, and working with technology that focuses on a complete product instead of supplying technology components. All these factors will not only push your business ahead of your competitors, but also move your manufacturing into the future.   M

View More