Scheduling in High-Mix Low-Volume Manufacturing: Why Rules Break

High-mix, low-volume manufacturing scheduling poses significant challenges for traditional ERP systems. These systems are designed for repetitive production and struggle when faced with the complexity of job shops, contract manufacturers, and specialty fabricators who produce hundreds of part numbers in small quantities. Factors such as sequence-dependent setup times, shared constrained resources, and dynamic arrivals create combinatorial complexity that standard rules cannot handle effectively.

Scheduling in these environments requires optimisation rather than just sequencing. While ERP systems provide a starting point, actual production sequences are often determined by floor supervisors who make daily decisions. The limitations of static rule sets become apparent when thousands of jobs and machines are involved. Shared bottleneck resources, sequence-dependent setups, and dynamic disruptions further complicate the scheduling problem.

Most ERP systems employ finite capacity scheduling or simple priority rules, which are fast to compute but fail to optimise across the entire problem. Advanced Planning and Scheduling (APS) tools offer more sophistication, handling sequence-dependent setup times and resource contention better than standard ERP. However, APS still requires accurate master data and struggles with frequent rescheduling in highly dynamic environments.

AI scheduling approaches, such as reinforcement learning, genetic algorithms, or ML-enhanced heuristics, tackle high-mix, low-volume manufacturing scheduling from a different angle. These techniques learn scheduling policies by simulating millions of scenarios, enabling near real-time rescheduling as disruptions occur. AI scheduling can deliver improvements like 10-20% reduction in total setup time, 5-15% increase in on-time delivery, and faster response to disruptions.

However, these benefits hinge on accurate data: reliable routing times, up-to-date setup matrices, and real-time machine status. Without quality data, AI scheduling may exacerbate existing problems.