Developing Lean cells in the so-called “fat” electronics assembly factory.

It has been years since the concept of Lean production came to PCB manufacturing. After making impressive strides in the final assembly stages, the tide seems to have ceased when it comes to SMT machine processes. SMT processes may be increasingly fast, efficient and productive by themselves, but overall, SMT does not perform well according to Lean principles. Is this a fundamental truth about SMT production, or is there another path to take?

There are two key elements to understand about the application of Lean thinking, one commonly understood but the other at times elusive. Lean, as the elimination of all waste, is the popular approach. The mistake, however, is that this should be regarded more as the end-result, not necessarily the initial approach.

Engineers relish the opportunity to exercise the tools of their training under the banner of “Lean.” Industrial engineering can be quite a fascinating subject because analyses are made of time and motion of a process to remove any action or operation that is not necessary. The simple idea of Lean aligns with this approach, as industrial engineering introduces more definitions of waste, in addition to the traditional focus on the waste of production time.

This approach to Lean, however, is backward. The starting point for Lean should not be the analysis of waste but, rather, the analysis of the commitment that the process is inherently making. This comes back to the second element of Lean, which, although elusive, is to many the most important: the “pull signal.” Any perfectly running process, without operational waste and regarded as “Lean,” producing output at the highest rates and quality, is actually a complete waste if the product that is being made is not required at that moment. It is a lean cell in a fat factory.

The “pull” of Lean distribution. If you look it up on the Internet, “cell production” brings up more pages on genetics than Lean manufacturing. The excitement of Lean seems to have died down. Links to business studies within university courses describe cell production in a simple way, such as the difference between a serialized production operation on a regular mass production line compared to parallel production using cells. Then, the emphasis goes toward discussion of additional needs of equipment and material supply. Again, the essential part of Lean – the pull signal – is missing.

The earliest consultants of Lean cell production took a completely different approach. Analysis started at the end of production to see and measure the way in which products were demanded from the factory day by day. It was clear that the production lines were not making what was being shipped, and that the warehouse was taking the strain in between, with days, weeks, or even months of stock being held.

Cell production was introduced with the idea to make in one day what was known to be needed for shipping on the next day. This could not be achieved by regular mass production lines because, for example, there were quantities of 20 different products shipped every day, but there were only five final assembly lines.
Unless lines would change over between products four times a day, the output requirement for the day could not be achieved. Bearing in mind the model changeover time of the day, the lines would be down too long, and output could not be completed in time. It was impossible.

The cell production solution, however, took the many operators from the serial production line processes to create, for example, 100 cells. The demand of the 20 products could now be split over the 100 Lean production cells so that pretty much the daily output could be made without any cell changing product. Each cell would make the whole product or at least a key subassembly. Issues around availability and sharing key resources such as testers of course need analysis by the industrial engineers. Keep in mind though that the output, and therefore throughput, is exactly the same in each case, so there is nothing fundamentally blocking this approach.

The Lean analysis of the cell operation takes over at this point, in its proper place at its proper time. The headline result from the analysis of the daily demand and the implementation of cell production was not the saving of operational costs through increased throughput, although this was recorded. It was actually the investment and depreciation cost saving of warehouse stock, which with Lean production had all but disappeared.

This kind of cost savings, as well as the ability to react fast to customer demand changes, is especially important in these days of Internet ordering and direct shipping. The benefits of cell production today can easily outweigh most other cost factors in manufacturing. This is quite an expected result when the full “value stream mapping” process of Lean is correctly applied, starting with the pull signals, and eliminating needless commitment.

Giving up on Lean SMT. That said, why is there still the resistance to Lean where SMT is involved? There are two points at which most attempting to make SMT Lean give up, often without a fight. SMT, a process relying mainly on expensive automation, traditionally works on a shift pattern that spans a full 24 hr. per day, to maximize equipment asset utilization. Manually intensive production operations tend to work only one or perhaps two shifts within the daytime.

Manual assembly is geared to processing pieces faster than SMT on a unit basis such that both remain generally in sync over a period of time. This brings the inevitable need for a buffer in between SMT-PCB‒related processes and final assembly. There is also the added complication that SMT has to be processed twice: once for each surface of the PCB and again for test processes that tend to require more manual operation, especially when it comes to visual inspection, fault diagnosis, and repair. The tendency is that because of the risk of SMT not being able to keep up in real-time with final assembly, the buffer between them grows and can represent many hours or even days of production.

ERP planning does little to help this situation because it does not understand the detailed nature of the SMT processes, treating production quantities as being organized into smaller and smaller batches, or in some cases using an inaccurately modeled continuous production plan where huge lead times are built in between processes to make it work practically.

These issues can be addressed, however. With an accurate planning and production progress tool using real-time event feedback from all processes, tracking SMT progress and completions can be possible. The buffer between SMT-related assembly and final assembly can be reduced to the working time difference, plus a very small buffer. This could then permit the pull signal from final assembly to travel up the production chain into the SMT area. However, this is where the second problem comes in.

SMT machine lines are streamlined high-performance machines, optimized with some of the most sophisticated tools seen in manufacturing. These tools take data that describe the products and work out the most efficient use of the machines to mount materials on the PCBs. Every effort is taken to remove waste of any unnecessary machine movements, additional transfer times, and needless machine down-time as one faster machine waits for a slower one in the line.

Here again, we find the excellence of industrial engineering making fast and efficient SMT processes, but which are often making the wrong products. The key reason for this is that the time taken to change the material setup on SMT machines between products is often the most significant loss factor in the SMT operation, especially where there is a higher mix of products. Optimization software for the machines has of course risen to the challenge, offering the ability to prepare a single common material setup across a range or group of products. Grouping works well in theory because the machine changeovers between products can avoid most or even all the physical changes in the setup of the materials. This was taken to an extreme by some, who decided to use this technique to completely eliminate all changeover time across all products by creating lines that could potentially build any product in any quantity, including on a one-by-one basis, at any time.

Again, the theory was great, the result, less so. First, optimization of the machines was severely compromised, as many more material feeders, and more machines, were required in the line to be able to cater for all of the materials needed over all products. The output performance of the machines was therefore very poor, down to a mere fraction of their full potential, bringing with it inefficiencies equal or greater than those that were trying to be avoided.

Second, as time went on, the product mix changed, with new products being introduced and end-of-life products being discontinued. The new products needed yet more feeder positions for new materials, while others went unused or were inefficiently reallocated. It had been so difficult to create the common setup in the first place that the impact of changing it would also be significantly loss-worthy in itself, and so the line was usually allowed to simply become less and less productive. This degradation issue also affects regular common material feeder setups.

The choice of which products to include in a common loading plan is today normally done while using the SMT programming tools, which show the effect on performance caused by that particular grouping choice. Grouping similar products has less effect on machine and line efficiency than grouping unrelated products where there are significant differences in materials, and quantities of those materials, used. This selection of products for each group ends up having little to do with what is actually going to be needed day by day for the final assembly processes and so is very unlikely to directly satisfy demand. The buffer stock and resultant disconnect of Lean SMT is therefore unavoidable.

SMT has been stuck in the way that it has been doing program optimization because it is a key task that can greatly affect the performance of the factory. The whole process needs to be turned upside down. Rather than start with the optimization of the SMT machines, line optimization, and then performing material grouping, start instead with the optimization of the material grouping based on delivery requirement, as closely coupled to the assembly process as possible, reading their demand as the pull signal.

For this to happen, a different approach to SMT planning and optimization needs to be taken. The final assembly production, whether based on lean cells or traditional lines, can be monitored as they follow their schedules, lean or otherwise. The expected call-off for SMT assemblies can then be calculated and set up as the delivery requirement. This will often be managed by the ERP or MES systems, or even in Excel.

The new generation of SMT dynamic production-planning software can then take these data, on a daily basis or even more frequently if required, and set them into context with what is already being produced by SMT, the feeder setups already in place, and build in any new or changed requirements. Optimization then takes place to find SMT grouping based on this, taking into account both the deliveries required and a simple but accurate modeling of the current machine and line performance, including times to change material feeder setups. In this way, the correct context, priority and weighting for decisions made relating product assignment to different lines, changes in setups, etc., are all considered together so as to produce the best possible and most efficient production operation.

This is different from what has been achieved in the past because it is reacting directly to the pull signal from the actual delivery need, ensuring that the minimum of buffer stock need exist between SMT, PCB assembly and final assembly. The final optimization of the SMT machines is performed as the execution of planning stage, rather than done before planning, resulting in no product allocation choices. Differences between the modeled SMT times and final optimized times will exist, but these should be less significant than the losses that have now been avoided.

Dynamic grouping drivers. Other changes related to the manufacturing operation work well with and support the Lean SMT operation. Implementation of Lean just-in-time materials logistics between the warehouse and shop floor is driven by pull signals of material requirements from both the machines directly and from the Lean production plan. Lean material logistics allow planning changes to be executed without the need of manual physical material reallocation and also help to retain material inventory accuracy, removing the vast majority of materials from the shop floor and saving significant investment cost.

Manufacturing engineering will also change, providing a matrix of the capabilities of the various lines against different products that ensures the planning system gets the maximum choice of product assignment. A single central machine vendor neutral programming system for SMT and test makes this easy to manage. The accurate and timely collection of normalized data from the shop-floor machines and processes enables the production plan to be more accurate and effective. All these factors can be managed by existing systems where capable, by manual methods, or by systems linked to or integrated on the same platform as the production plan tool.

SMT then, can be a Lean process, specifically from the perspective of the whole value-stream mapping. Achieving this brings the benefits of increased machine operational performance, while at the same time is more responsive to demand changes, and brings a reduced inter-process buffer requirement. Additional savings come from the extension of Lean operation to materials logistics and other resource management. The main change is that the optimization strategy for the SMT area has been turned on its head. Optimization is now performed from the perspective of requirement rather than capability. With all the complexities being calculated and managed by the production plan optimization engine, it becomes a matter of setting relative priorities for the business in the setup of the production plan tool, and then connecting the sources of information for delivery requirement, materials availability and current production progress status.

This approach can improve SMT even where the next-stage assembly lines remain traditional. The actual benefit in SMT productivity will vary depending on product complexity, the number of variants, product lifecycle expectancy, and the capabilities of the different machines and lines. Applying Lean tools for planning and materials enables companies that have in the past enjoyed high-volume stable manufacturing to retain performance levels, even when faced with increasing mix and production variation.

Figure 1. “Lean” applies to all levels of production, bringing improved efficiency to SMT.

Michael Ford is marketing development manager, Mentor Graphics (mentor.com); michael_ford@mentor.com.