Guest Author: Aniello Pepe, Director, Industry Solutions Group, Oracle
In my previous blog post, I highlighted how artificial intelligence (AI) can be applied both to do the same things better as well as to do things in a new way, with some examples on how that may apply to manufacturing. Here I wish to explore the concept a little further in the field of manufacturing execution.
The Planning and Scheduling Challenge
The goal is clear: using available manufacturing resources and available materials to achieve the target production plan in the most efficient way possible, in term of quantity, quality, time, and costs.
The classical approach is typically prescriptive: given the known constraints and forecasted status and availability of the production means and supplies, a detailed schedule is created applying various techniques and algorithms to state the exact production sequence to apply, i.e. determining the sequence and routing of parts in work centers on the shop floor.
As any strategist knows, no plan ever works as expected. Unexpected events can disrupt the best plan, and they often do. Indeed, sometimes the intrinsic complexity of the manufacturing context simply doesn’t allow you to take into account all possible constraints at the appropriate level of detail. Then plan recovering starts.
How AI Can Help Plan Recovery
Recovering from a disrupted plan is the daily nightmare for planners. An overwhelming amount of information about the running production environment needs to be taken into account in order to adjust the plan. At the same time, the impact of any change needs to be considered very carefully.
AI techniques can be of great help at this point. They can be applied to go through all the available information from different sources, including IoT data gathered from machinery on the shop floor, to create recommendations for a course of action. Simulation via digital twins of resources and production lines, predictive analysis for maintenance, and what-if analyses against multiple parameters are just some techniques for intelligent decision support.
Which Approach to Take?
Of course, it depends on the maturity of your process and the set of tools you rely on. As usual, the safest approach to innovation is to start small and grow from there.
I suggest making a soft schedule and constantly review it – in a continuous adapting mode - according to the indications provided by an AI-powered model of the production environment, taking into account all information from IoT sensors, digital twins, execution status and more.
The Next Evolutionary Step
A more aggressive approach is possible that relies more on autonomy and edge computing – moving more intelligence into the cyber-physical objects. It requires transforming your way of thinking about the planning and scheduling problem from the traditional top-down, prescriptive approach towards to a new bottom-up, self-adaptive one. This would entail giving up the development of a global plan and instead allowing shop floor resources to compete with each other to obtain orders to be worked. This relies on the concept of autonomous agents, representing both semi-worked parts and production resources, each competing to be worked or to get works to do and governed by the goal of achieving target KPIs (cost, time, quality etc.) at the best level.
The sum of local optimal results may not actually result in a global optimal result, but rather a sub-optimal one that constantly adapts itself to the actual shop floor conditions. This, however, is a far better result than a perfect plan that cannot be implemented and must be reviewed whenever there is any deviation.
How Oracle Can Help
Oracle is investing in AI-powered apps for manufacturing to enable you to move up the intelligence maturity curve towards manufacturing operational excellence. For more information about Oracle’s AI approach, check out our AI Applications for Supply Chain and Manufacturing. For more insight into how digitization is impacting manufacturing innovation, view Adapting Manufacturing for the Digital Age.