Manufacturers are rushing to keep up with the latest technology trends and perhaps the most significant ones are around the smart factory. Whether you call it Industry 4.0, Smart Manufacturing or the Industrial Internet of Things (IIoT), what all these initiatives have in common is the desire to maximize value from manufacturing data and improve overall manufacturing efficiency.
With the average net margin for the automotive industry at 2.07%, it’s not surprising that Automotive manufacturers are jumping on the bandwagon and are utilizing IIoT in order to produce quickly, cut costs, and run their daily operations smoothly.
Much of the industry’s focus is on the machines doing the manufacturing. How can we make machines more efficient, predict failure, perform maintenance just-in-time and maximize up-time? IIoT vendors explain that by taking sensor data and machine logs and applying advanced analytics, it is possible to predict failure and proactively fix issues before they impact production. Using this technology enables us to answer questions like “when does this machine need maintenance?” or “how can I make my line more efficient?” This is one of the core promises of smart manufacturing.
However, by taking the IIoT paradigm even further, it is possible to create even more value from data and analytics, drive extensive improvements in product quality and reliability, and extend the benefits from the manufacturer to the brand owner.
When using the term “Industrial Internet of Things”, the “Things” we are typically referring to are the machines. But what if we use the word “Things” to refer to the products that are being manufactured? What if we take the data from all of the machines and supporting systems used to build each individual product and connect it together using a digital thread?
We can take the configuration parameters, sensor readings, bill of materials, rework data and test and inspection data from each individual product as it moves down the line and create a digital “DNA” of the product. Even better, imagine what you could do if you connected the data from all of the components that comprise the product into this digital thread.
Applying artificial intelligence to this data enables us to answer many additional questions. For example, what is the true quality of the product? How likely is it to fail when used by a customer?
More specifically, it allows us to determine the root cause of issues in the manufacturing line and resolve tough engineering issues by understanding the complex interactions between components, consumables and the assembly process.
For example, what is causing parts to fail at inspection? Are my failures related to a specific supplier of raw materials or perhaps to a configuration on one of the assembly machines? Which of my suppliers provides components that provide my product the best possible performance?
To answer these kinds of questions, any IIoT or Big Data Analytics system needs to be built from the ground up to focus on the product and not just on the machines. According to a recent study conducted by IBM and Oxford Economics, manufacturers need to put IIoT products in place in order to mitigate the possibility of product inefficiencies, faults, or damages.
IIoT products tend to use time-series databases to store and analyze sensor data. These databases are great tools to resolve common time-series problems but they don’t help describe the relationship between data points collected along the timeline and the specific products that were going through a machine at any given time.
To reconstruct the DNA of a part, we need a way to efficiently tie sensor readings taken at the time each product was passing through the machine to the product itself using timestamp data.
Let’s take a welding machine, for example. Some processes use welding to attach parts to each other. As the welding probes bond the parts, variations in temperature or pressure can impact the quality of the weld. Understanding the complex interaction between the various physical phenomena as the parts are bonding is key to guaranteeing consistent welding quality. But you can only understand the interaction if you tie the data collected for each part to the detailed inspection results using a unique part ID and use artificial intelligence and machine learning to comprehend the data.
So what does it take to create a product-centric IIoT analytics solution?
First, it requires a data engineering exercise where all possible data sources are connected to a central repository and the data is cleansed and organized in a way that suits complex analytics. Product and part identifiers need to be aligned so that the complex relationship between parts and products can be comprehended by the system. The entire process needs to be governed by tracking each piece of data from source until it reaches the central data repository and preventing data from going astray.
Companies often underestimate the complexity of performing this task and end up with large datasets that are difficult to connect, understand, and analyze. Many CIOs report that they have invested heavily in building manufacturing data lakes but that users find it difficult to leverage the data and justify the investment.
Secondly, the data platform itself must be able to ingest considerable volumes of data from many disparate data sources and provide users with tools to quickly and easily access this data. This must be done while considering the genealogy of each product and the complexity of the manufacturing process.
Also, since IT departments are already struggling with a profusion of systems and technologies, the data platform must “play nicely” with other existing systems and be able to leverage data without duplicating it.
Thirdly, it needs to provide the right combination of out-of-the-box solutions to common manufacturing problems as well as do-it-yourself analytics tools. All too often, IIoT platforms focus on the latter, providing just building blocks which require significant IT and data science effort to connect and build solutions. This means that it takes a long time until an IIoT solution starts returning the heavy investment it required.
For really complex manufacturing problems, the platform must connect to the latest and greatest AI tools so that data scientists can build and train machine learning models. Data scientists expect to be able to use the tools of their choice and not be tied into the tools and libraries provided by one specific vendor.
These models are then enabled and deployed on the edge so that action can be taken in real time. For example, for a machine learning algorithm to categorize the quality of individual products and prevent bad parts from shipping, it needs to be connected to systems like MES to manage the flow of material.
If done correctly, the value from investing in these IIoT solutions and analytics platforms can result in higher sales margins, larger profits, and thus, a greater ROI.
Creating a product-centric IIoT solution is a complex but highly rewarding task, with customers reporting quality gains of 50%, efficiency improvements of up to 20%, and significant acceleration of new product ramp thanks to the ability to quickly analyze the root-cause of issues as the assembly process matures.