Many manufacturing organizations are driving value through automated visual inspection, and have embraced how and why DataOps platforms become a critical success factor as they enable accelerated contextualization at scale in support of tailored decisions.
The cost of quality in the production and manufacturing process is high, amounting to often double-digit percentages of annual sales revenue. This is why the concept of quality inspection emerged about a century ago as an important process across the value network. It is now a critical process for industries that need to ensure brand standards for their customers. It’s what ensures the safety of your glass, the uniformity of your favorite cookie, the perfection of the steel used in machinery, or even the particular shade of red in your lipstick.
Because there is no “Undo” in manufacturing, quality control has needed to come a long way since the dawn of the industrial age – from people on the factory floor, relying on their own attention to detail and high degree of alertness, to where we are today, at the digital dawn of the manufacturing industry.
We have the capabilities today to capture images needed to perform quality inspection, rather than risking human error in the process. This used to be called “visual inspection”, but now that it’s powered by computers and machine learning, it is referred to in the industry as “computer vision inspection”.
Slowly but surely, manufacturing companies are making the transition, from manual to computer vision and autonomous inspection, to verify that all goods and processes meet the same consistently high standards. Computer vision is what has enabled greater consistency across production and manufacturing. It serves as an important early step towards more data-driven operations, but it’s not the end of this story.
To capture business value, computer vision is an important starting point
Let’s use the example of lipstick manufacturing. Imagine that today, “S’il Vous Plait” red 124 is on the assembly line. In this process, the manufacturer is concerned with several things: the efficacy of the lipstick itself, the safety of the product and raw materials used, and of course, the consistency of the color and shine. Failure in any of these areas will result in a product recall, consumer dissatisfaction, and potential damage to reputation.
Throughout the lipstick’s journey, computer vision is what has been capturing video or still images along the entire production lifecycle. Artificial intelligence and machine learning models analyze images and detect anomalies, such as a faulty tube or discoloration in a manner that far surpasses the capabilities of the human eye. This itself is not entirely revolutionary, but what is revolutionary is this: the manufacturer is able to extract further value and insights from the visual data that’s being captured, and can further scale that across other parts of its operations.
Now, what if this “faulty” tracking is happening at specific stages of the production line, from incoming supply goods to finished products, and is reconciled to the specificities of the bill of material and bill of process recommendations for a specific production site? What if this will help drive predictive quality controls processes where digital operators are proactively adapting production parameters based on machine- or system identified deviations? What if this can also help track defects from the suppliers of the lipstick casing? There are endless possibilities.
Innovation in how we use data, turn it into insights, and then make data-driven decisions, is just beginning to enter the manufacturing realm. Companies are now confronted with vast amounts of data, including an increasing amount of visual data, though many struggle to understand what to do with it. Figuring out how to put that data to work is the next inevitable step in their transformation. By storing, contextualizing, and integrating that visual data with the data set of the entire production, manufacturers are able to both map and query data from their entire history of visual data and glean deeper insights into every step of their operation.
Data drives real-time decision making across operations, across factory floors
So, what does this mean in practice? If we want to search for and compare the entire history of discoloration in “S’il Vous Plait” red lipsticks, we can pull up the image history and analyze the differences between the discolorations as well as any anomalies or similarities in the entire production process that may have led to this quality error. The image data is searchable. Everything produced is traceable. And it’s integrated into the entire data ecosystem so that you can better understand what went wrong, where, more importantly, why, and ultimately how to address it.
And that’s when things get interesting.
When you know what happened in a process or how things could have been done better, you suddenly have an entirely new tool at your disposal to make better decisions, faster than ever before.
This is just one component of immersive decision-making in operations, and much more can be considered by combining capacity utilization, predictive maintenance, predictive quality, and energy consumption. Enabling multi-criteria business optimization on the factory floor becomes a vehicle for operations’ top-line and bottom-line improvement, talent growth, and retention.
This is what is referred to as real-time decision-making based on actual, operational data – gathered and contextualized from your production line at any given moment. This enables you to easily detect pain points in operations and enable workers to simplify and optimize based on data – making decisions that can improve performance, cut costs and minimize environmental impact, waste and emissions.
Empower all your workers with data, and crucially, with insights
This is hopefully meaningful for those sitting in the boardroom, but even more so for managers and for the people on the frontline, the ones managing production or handling data that enables them to make real-time decisions. To these types of workers, data-driven insights that empower real-time decision-making make all the difference. It’s like a living encyclopedia at their disposal; one that records past and present on the line so that the employees can direct better outcomes for the future.
We can make all the slides we want about digital transformation in manufacturing. But in parallel with the deployment of digital supply control towers, the real transformations are also happening today in the place where the value of digitalization is revealed: the operations. That’s where real time data and insights can be put to work to reap real value and make smarter decisions so that at the end of the day, the consumer can trust that their lipstick will always be the perfect shade of red.
Slimane Allab is an industrial engineering, consulting, sales, and business leader with extensive experience in manufacturing, industrial technology and management consulting. Today, Slimane leads Cognite’s manufacturing vertical, helping global manufacturers leverage data and scale technology to build agile, future-ready systems in support of their current and next generation operations. Slimane has a PhD in Industrial Engineering from the Institut polytechnique de Grenoble.