Is it OK to build to a schedule?
When we think of “Lean” we think of the analysis and removal of all waste from a process or set of connected operations. Often though, unless considering the factory operation as a whole – however that may be defined – going to the nth degree to remove waste from a particular operation reduces flexibility and may even contribute to “unavoidable” losses elsewhere. If we were to consider “Lean” for the whole factory operation, how would we approach it, and what form of Lean application would a solution look like?
The focus of flexibility, also a key requirement for Industry 4.0, is to be able to dispatch products in the required quantities on short notice. This can be achieved by simply having a huge warehouse full of whatever products and variants may be called for, of course. Although this approach has a huge cost of ownership and management, with significant risks of product price depreciation, it is used to a greater or lesser extent in the majority of SMT-based manufacturing operations worldwide. The associated costs become a part of the manufacturing cost of the goods, one way or another.
The Lean approach to achieve the same goal avoids the need for substantial finished goods stock, which many would then regard as effectively being “build to order.” An SMT-based operation is somewhat challenged in this respect because, with the nature of today’s SMT processes, productivity is inversely proportional to change. The unloading and reloading of materials at changeovers between work-orders of different products can represent 50% of the total machine available time where there is significant product mix, and even represent as high as 95% or more where batch sizes are very low and there is a high mix of products.
Solutions to reduce material setup time by creating setups that can support several products without the need for making changes certainly helps reduce this loss time, but at the expense of reduced efficiency in the placement program. The losses are effectively moved from one place to another. The use of removable feeder tables can also help reduce the material changeover time, but again this introduces cost elsewhere because the amount of materials stored on the shop-floor increases dramatically. Although these approaches help, they are not solutions to the SMT flexibility versus productivity loss issue. Substantial stock of subassemblies that have gone through the SMT processes, as well as completed products, are still needed in most cases, with SMT processes overall failing to become Lean.
It’s time then to focus on the application of Lean to the SMT factory as a whole rather than getting bogged down at the process level. An easily understood analogy is to consider a typical trip for the average traveler through an airport. Consider how much of the total time taken from arrival at the terminal to flight departure is of added value to the travel process. It is similar for a product being created in an assembly factory, from the first raw material, likely to be the bare PCB, to the dispatch of the product to the customer.
At the airport, the first challenge is check-in. Normally it can take some time waiting in line, if only to drop off a checked bag. The added-value component is the time spent interacting with the person at the desk, perhaps representing two minutes after 20 minutes in the queue. Then there is the walk to join the line for security check. Following that, we have the time in line to check your ID and boarding pass, the check taking perhaps only 10 seconds; then there is the body scan and the scan of carry-on items, perhaps a one minute value out of another 15 minutes waiting. At least the next long wait in the departure area affords the opportunity to create some value checking emails or getting something to eat, but these also are not adding value from the travel perspective. Even after the boarding process, there is the long taxi out to the runway, again in a queue, before takeoff. All in all, what is for many travelers about a two-hour ordeal, the actual added value of the airport experience, including the walking from entrance to gate, is only about 10% of that time.
The same is true for product that goes through successive production processes. The actual time spent being mounted, tested and assembled is probably far less than 10%, or even 1%, of the total time the product spends “in process” in the factory. When it comes to flexibility, the overhead to implement changes to the flow of products can be a hundred times more than if products were flowing through the processes without waiting.
Imagine the scenario where inter-process and final product WIP are not permitted, so the factory makes each product without any waiting time. Each product unit moves immediately from one process to the next for immediate completion – starting as a bare PCB, then being marked, going through the screen printer, then perhaps solder paste inspection, a line of SMT machines, automated visual inspection, then turned over and the same sequence applied to the other surface. PCB manual assembly is performed next, then on to in-circuit test. Other subassemblies required for the same final product unit are made in parallel so that each required subassembly is available at the same time for final assembly, functional test, packing and shipping. The whole manufacturing process should be completed in about 10 to 20 minutes. Running one unit of a product through a factory in this way presents few issues because every process is ready and waiting.
The problem, however, is that during this time most processes in the factory were idle, and the waiting was shifted from product to process. Multiple products must be able to be made at the same time to regain productivity. As the number of products to be made concurrently rises, the utilization of machines and processes increases, but inevitably leads to waiting time as bottlenecks occur at key processes, especially as very frequent changeovers of materials are likely on the SMT machines. If the SMT shop floor is left to run on a pure kanban system, the result in productivity will be poor indeed.
The solution may lie in something that normally is considered not very Lean at all: a schedule; however, it is finitely optimized for both the delivery requirements and modeling of the SMT machine processes, including changeover time. This finite schedule optimization fully considers the availability of materials and resources, which ensures every decision made by the optimizer will be executable without issue. Consideration is also made to reduce the need for in-process WIP and finished goods storage.
For SMT, there is one further crucial step to take. In addition to being able to model the operation and changeover times of SMT machines with reasonable accuracy, the production plan software must also be able to generate dynamic common setups of materials to reduce changeover time. Exploration of all optimization opportunities to group products together must be done together at the same time as the main product flow sequence optimization to truly find the best possible paths for products and their associated work-orders.
The nature of customer demand may have become much more immediate and volatile, but it does not need to focus on a single piece “make to order” basis. Delivery demands remain as a flow requirement to the factory over periods of days, or perhaps a week or more. This retains the scope for a short-term planning system to be possible. The finite optimization engine with built-in support for SMT modeling and dynamic creation of common material setups is an aid to shortening the time of products through the factory, while decreasing waste from a WIP perspective, maintaining production efficiency, and providing increased flexibility.
is marketing development manager, Mentor Graphics (mentor.com); firstname.lastname@example.org. His column runs bimonthly.