AI has started to touch every industry, with market leaders already gaining an outsized advantage. Manufacturing is no exception. Factories that have operated with the same stable, traditional processes over the past 20 years are being superseded by flexible processes with data driven software at their core.
Reinvigorating these processes comes with significant hurdles, so before doing so, it’s important for manufacturers to clearly define and consider their catalyst for this change. Having this base understanding will help organizations overcome production challenges and see the real impact of this technology all while staying true to their overall corporate strategy.
1. Fragmented Data
Effective AI integration requires high-fidelity data combined with contextual knowledge. For factory floors that have operated the same way for decades, much of this critical knowledge lies within tribal know-how and comes from years of onsite experience.
For example, an operator might be able to walk by a stamping press and know if a part is in specification. The first hurdle manufacturing teams will encounter when implementing AI is being able to link all the data produced by these processes to create a full digital narrative.
This narrative quantifies current competitive advantages that have allowed an established facility to achieve its longevity.
A good starting point is to begin collecting data from a non-intrusive, observational perspective in an area where the company has a competitive advantage. Starting a data collection pilot on a machine that is stable, predictable and well-understood, allows companies to create a reliable baseline with consensus.
These applications lay at the crossroads of capital allocation, material importance and vast wells of tribal knowledge.
Take inventory of all the independent variables possible from that process such as feed/speed or temperature, with variation that can be captured and translated into tighter production tolerances. Capturing these variations can be tedious but can be streamlined by utilizing an edge data device to collect all machine data.
It’s better to have excess data than risk not having enough. This helps paint a clear process narrative for cross-department teams to align on key metrics to be constantly monitored and that largely impact the quality of finished goods.
Record all the data possible now as it will be an invaluable asset to training models. Think of this data collection as an investment for a factory’s operational improvement future. Synthetic data can fill gaps, if necessary, but utilizing as much real machine and organizational data as possible will be more impactful, and doing so requires imminent collection.
2. Lack of Real-Time Visibility and Closed Loop Integration
Collecting data and quickly sharing it is critical but having systems to analyze and respond to findings at speed enables manufacturing teams to see the real impact of their work.
Data can become siloed and stagnant in IT systems without an environment to influence operational technology (OT) actions. Real-time process tuning requires relevant data without latency.
Without a control architecture that can keep the speed, accuracy and precision of these machines, recommendations for improvement can come too late to implement. This can cause lost production or quality issues.
To prevent this, manufacturers can close the loop between data collection, analysis and action. Focus on improving the speed of information flow between process control, such as a PLC, and data analytics software to enable faster decision making without modifications to the core control logic.
On-edge dashboards and time-synchronized video with data playback can help validate insights during system changeovers and operation. Using independent control processing units (CPUs) allows teams to separate real-time machine control from higher-level analytics, enabling safe and governed mediation.
This helps to create a closed-loop system where data informs and enables timely action, first by operators and engineers then eventually by automated systems as trust, internal governance and maturity grow.
3. Culture
Beyond technical challenges, organizational culture can either be a significant barrier or an accelerator of integration. Those who have worked on the factory floor for years can be skeptical of change, and rightfully so, as the existence of long-standing facilities is proof that teams have not been swayed by technical fads but instead focused on their competitive advantage.
It’s important to include all stakeholders in these updates starting in the data collection phase by being intentionally transparent. Instead of leading with the technology updates, teams should first identify current operational challenges on the factory floor.
This alignment on current challenges helps create a clear explanation for the shift to a data-first culture and gives merit to how these changes will help drive progress on the factory floor. Outline the process for change, what benefits could be found from these advancements, how teams will be a part of it and how they can shape it based on their input and insights.
These steps are critical to ensure teams are confident in the rollout and can share ownership of success.
Utilize the data from an initial pilot as a proof point to help teams become more successful and recognize the software as a tool to help augment their roles. Including employees in AI integration from the start helps build trust, accountability and buy-in for these new tools and processes.
It also gives leaders access to diverse perspectives across the factory floor with visibility into any concerns surrounding the technology and workplace culture. This approach gives teams historically not privy to emerging technology, who are often most skeptical of change, ownership of their future and input on where advanced process controls should be implemented based on ROI.
Laying the Groundwork for the Industrial AI Future
As AI capabilities rapidly evolve, new hurdles will arise. Factories with stable, repeatable and in-demand processes can find competitors surpass them with advanced process controls.
Just as the best time to plant a tree is years ago and the second-best time is right now, a data-first factory floor becomes advantageous in the future.
Creating a basis of employee inclusion, strong data collection and closed-loop decision making while choosing process automation tools that can be compatible with future innovations, will be the winning formula for factories for years to come.
Thomas Kuckhoff is the senior product manager at Omron Automation.
