Resilience has been one of the biggest topics in mechanical and plant engineering in recent crisis-ridden years. However, this does not mean simply persevering. Appropriate tools are needed to move forward: namely AI and intelligent algorithms for detailed planning.
Unforeseen but completely ordinary disruptions such as bottlenecks on machines are familiar to almost all manufacturing companies. Production planners in single-item and small-batch production must then be able to make good decisions quickly, despite complex and dynamic processes, as customers tend to measure the quality of a supplier by whether they meet the deadlines they have promised.
See also: Gen-AI’s rewards are clear, but the data must be ‘clean’
Using AI-based production planning, throughput times can be reduced by up to 62% and productivity increased by up to 11%, a study conducted by INFORM showed. Such systems achieve these results because they take more influential variables into account than a human planner ever could.
After all, the countless interdependencies of individual production steps for preliminary, intermediate, and final products—as well as the resulting restrictions in terms of personnel, material, and machine capacities—are impossible to keep track of concurrently. Even 10 production orders could conceivably have 3.6 million different sequences. This is not even considering various priorities, capacities, throughput times, or other relevant planning parameters.
Keep in mind, in theory and hopefully in practice, planners must factor in considerably more input variables from existing orders, available machine capacities to workforce vacations and unexpected time out, as well as setup times and transport routes in the plant. What if a spare part or rush order must be produced on short notice? You don't want to be the person who must work it all out. This is where AI comes in.
How APS systems work in plant decision-making
Classical ERP systems plan against unlimited capacities. For example, they determine the lead time of individual orders, but cannot dynamically distribute competing orders among available resources. The algorithms of an APS (advanced planning and scheduling) system, on the other hand, calculate a sequence that aims for the best possible overall result.
These solutions draw all planning-relevant data from the ERP system to generate the corresponding schedules, taking into account available resources and capacities. When examining a mathematical decision model, the AI—to put it simply—continuously performs a mathematical proof that shows that the best solution for a problem must lie in area A and not in area B of the solution space. In this way, the algorithm gradually approaches the best possible production plan.
See also: Industrial Applications of Generative AI: A Smart Industry eBook
In doing so, good algorithms always suggest these decisions with the entire production system in mind. For example, the calculation might show that more rush orders can be served on time overall if another order is delayed. Machine allocation is therefore not just a matter of scheduling the next follow-up order on a free machine. Instead, AI might recognize that it could be better to let the follow-up order wait to free up a machine that would be more suitable for the rush order.
If AI recognizes delivery deadlines can only be met with overtime and extra shifts, it will issue an alarm and a clear planning forecast well in advance. If it recognizes material bottlenecks, it will inform the procurement department ahead of time which parts must be procured by which deadline. As AI-based algorithms can identify patterns and draw conclusions from them, they calculate lead times based on their findings, instead of using the date notified by the supplier for a particular order.
Digital production planning makes a good target for investments
It has been found that weak points identified by certain algorithms not only help AI achieve better planning results, but they also show the company which orders justify new employees, whether it’s worth investing in additional machines because capacities are always tight on existing production lines, or which suppliers the purchasing department should reward for regularly exceeding delivery deadlines.
See also: Process shortcomings a high hurdle for phasing in AI, survey finds
However, for digital production planning to develop to its full potential, all departments involved in the manufacturing process should be involved in its implementation. After all, departments such as procurement, design, and sales should work hand in hand with manufacturing.
To achieve the best possible results, the foremen and machine operators from production also need to be at the table. This is because they often have access to valuable knowledge and experience that isn’t mapped in data—for example, line setup processes and the time required to start production on the individual lines. This means that AI and the specialist knowledge of the foremen can be combined to achieve great advantages.
Anyone who takes this to heart can strengthen their position within the market with AI-based production planning. This ultimately allows companies that deploy AI-based production models to deliver on their production lead times and to satisfy customer orders within the committed delivery dates, even when unexpected disruptions occur.