It was 1989, when a Fortune 500 company was building a technological marvel. It was one of the largest FMS projects in the country. At the time, the idea of a flexible manufacturing system (FMS) was the dream of executives, engineers, marketing people, and definitely programmers and systems analysts.

Computer numerical control (CNC) machine technology got its boost in the '60s and '70s. More and more machines were built using CNC technology, enabling substantial improvement in quality, machining capability, and process flexibility. This technological breakthrough enabled companies to perform at standards never achieved before.

However, the change did not come without a price. The CNC machine's cost jumped by an order of magnitude compared to conventional machines. The workforce had to be retrained. Maintenance people needed higher skills to deal with the complex electronics. A new function – programming – was added. Engineering departments had to add people, computer hardware, and software.

CNC technology was probably responsible for the major shift in companies' cost structure over the past two decades. One CNC machine tool replaces several conventional ones, resulting in fewer operators on the shop floor. The ratio between factory labor and total overhead cost has shifted from a typical range of 1 to 2 to 3 to 10.

FMS was born to capitalize on CNC technology: to link a series of machines that could start with raw material and finish a complete part.

Beyond that, FMS was to provide the ultimate in flexibility: to perform at near zero setup and changeover time. The vision was to bring the economics of a mass production line to the manufacture of a batch of one.

In theory, it sounds good. In practice, more often than not, it falls short. That's what our Fortune 500 example company found out too late.

The FMS was designed to handle the manufacture of up to 300 different parts. It replaced a line of CNC and conventional machines in a multi-batch process flow. Although the existing line was operating satisfactorily, the hope was that the FMS would reduce cost and improve productivity. Performance was far from satisfactory. Although the FMS was operating three shifts around the clock, output averaged only 50%.

What went wrong? The answer lies in the difference between theory and practice. As in many good ideas, the problem does not lie in the idea itself but in its execution.

FMS is a closed system where several CNC machines are linked through a transfer medium (like conveyors) and a closed-loop feedback system. The feedback system includes data flow relative to scheduling, banks of machining programs, tooling, material being processed, quality control points, and much more. In short, it's an extremely complex system.

There are hundreds of variables and different elements that have to be accounted for in making a product. That complexity is indigenous to the way any FMS is built – almost every component is dependent directly or indirectly on the other. Therefore, any disruption of a component's performance (and granted it will happen) impacts the total system's output.

When the FMS system was designed, very sophisticated simulation software was used to approximate downtime. However, day-to-day disruptions, which were not considered, eventually created chaos. These disruptions could involve material quality, a part that does not match the schedule, a broken or dull tool, computer downtime, programmer errors, or a conveyor line problem. The list is very long. Because of their randomness, no simulation could be expected to predict their cumulative impact.

The problem is rooted in the way we build the FMS. It is a close-coupled system where every disruption in one component virtually shuts down the entire operation. Disruption of this nature may not be large, but when you accumulate five minutes here and ten minutes there, it comes to a significant loss.

What we need is to decouple the FMS – we need to create the buffers between its key components that will be able to absorb the system's fluctuations. Buffers could be in the form of material positioned between two machines, standby spare tooling, space buffers, or a transfer medium that is not linked physically to machines. By taking this approach, we can capitalize on CNC technology to form a truly flexible system.