Is a high-mix environment suitable for a pull signal setup?

As one of the developers of the first Lean, just-in-time (JIT) material logistics systems for electronics assembly back in the early 1990s, I can say this genuinely “smart,” computerized, Lean technology was a revolution. Over 25 years later, however, we see the majority of companies every day getting into a huge mess with their incumbent “push” systems, with bloated kits of SMT and other poorly managed materials on the production floor. What is the problem here? Given the adoption of Lean materials can reduce material-related costs in assembly manufacturing by more than half, why haven’t all factories adopted Lean material management? Let’s look at the true state-of-the-art to unravel truth from myth, and set out the practical requirements to make it happen.

At the heart of JIT Lean materials is the generation of the pull signal. This is where the machine or process has to send a signal to the supply-chain logistics to request replenishment of materials, so needed materials arrive before the existing supply exhausts, “just in time.” To do this, computerization is required to use data from different sources in real-time and make decisions. Data about materials and their consumption first need to be processed.

For the sake of convenience, let’s consider the case of a reel of SMT material, though the same process applies across all types of materials that may be consumed as part of the production process. The first step is to understand the number of materials on each individual reel. As materials are booked into ERP at receiving, each reel needs to be uniquely identified and a record made of the initial number of materials. With materials set up and verified on the machines, the count of materials remaining on the reel can then be decremented as they are consumed.

The consumption signal can be derived in different ways. A simplistic approach is to understand when a PCB has been processed, either through a machine signal, button, sensor or barcode scan, at which point the count of components of each part number from the bill of materials for the PCB at that process is decremented from the reel quantity by the computerization. The issue with this approach is the machine may create material spoilage that also needs to be decremented from the reel. In the case of SMT, machines with more modern interfaces provide this information either as a series of events or as a traceability record for the PCB.

In a perfect system, all spoilage would be accounted for and decremented accordingly; otherwise, the reel of materials may exhaust sooner than expected. An example of unaccountable spoilage is the waste of material when putting the reel on a feeder. An allowance or assumption for this can be built into the computerization logic if required. In cases where it is not possible to gather all spoilage data, an allowance should be made so materials are pulled sooner than otherwise required to compensate. Ultimately, exhausting materials creates the opportunity to “self-correct” because unaccounted materials in the computerization can be assumed to have been spoiled.

In cases where splicing is adopted, a signal from the machine to state the splice has been detected on the feeder is acceptable for an exhaust message, meaning the machine itself does not have to stop to self-correct the material count.

Pull Signal Generation in a High-Mix Environment

A common myth about Lean materials is they are not effective in high-mix scenarios. If we were to consider only the raw pull signal as described, which clearly is effective only for material replenishment, this would be true because it only supports continuous production. The Lean materials computerization has to be far cleverer to modify the pull signal without losing the “Leanness” of the solution.

The key addition, then, is to manage material pulls between work orders, bearing in mind different use cases, depending on the equipment and operational methods. When work orders are completed, a changeover is required, which can mean the changing of some or all the materials set up on the machine. There is no point pulling materials toward the end of the work order that will not be needed.

In the high-mix environment, the pull signal needs to be modified in line with the production plan. The Lean material computerization needs to know how many PCBs will be produced in the work order and maintain a count of how many PCBs are remaining. Once enough materials are on the machine to satisfy the remaining production, the pull for that material can be stopped.

Then, there is the issue of requesting materials for the next product to be made, which needs to be physically at the process just before the changeover. The pull signal for these materials needs to be started accordingly, with consideration made for materials that may be common to both the incoming and outgoing work orders, because there is no point returning materials to the warehouse that could continue to be used on the machine. The changeover time can be greatly reduced, even in cases where common feeder setups are not used, if the whole line material requirements are considered, rather than on a machine-by-machine basis. The computerization needs to provide clear instructions during the changeover procedure to remove and return, set up and verify, leave in place, or move and verify materials.

In high-mix scenarios, additional trolleys are common, which may be set up with materials for the next work order in advance of the changeover. The logic within the computerization needs to understand when such trolleys are assigned and modify the pull signal generation logic accordingly, including knowledge of pre-existing materials on the trolleys. In general, unused materials should be returned to the warehouse or other storage areas, such as local towers, so they can be allocated to other work orders, saving space on the shop floor. The count of remaining material quantity is remembered by the computerization, so no counting of remaining materials is necessary.

lean1
Figure 1. Shop floor material, before (left) and after automated material pull and logistics.

From “Pull” to “Pick”

The choice of materials to pick from the warehouse is the next consideration for Lean materials computerization. As well as the usual FIFO (first in, first out) logic, choices of materials need to take into account restrictions such as specified or consistent vendor, engineering preference, or substitution, moisture-sensitive device (MSD) status, LED binning, and existing feeder loading. In addition, material quantity needs to be checked to ensure enough materials are picked to satisfy the JIT cycle time, especially as partly used reels may have few materials remaining. Also, the material consumption rate can be high, meaning that, again, more than one reel may need to be picked.

The JIT cycle time must be set to be greater than the time needed for logistics to complete the whole cycle, from taking the pick instruction, delivering all materials to the shop floor, picking up returns, and putting them away in the warehouse. This can typically vary from 20 min. to an hour, depending on the relative locations of the lines and warehouse. As long as there is enough logistics manpower to physically satisfy production demand, a practical JIT cycle time calculation will always be possible. In some cases, intermediate material storage locations closer to the line may be preferable, in the form of storage towers, for example. The time to pull materials from these and to store partly used materials is far less than from the warehouse.

Additional Kanban signals are needed to pull new materials from the warehouse to replenish the storage tower if new material storage in the local towers is required. This results in a marginal increase in material logistics effort and an unnecessary increase in material stock because the towers will likely be stocked to provide enough materials for any expected work order. This becomes a hybrid “pull-push” system, significantly less Lean than the pure “pull” and far less flexible for high-mix because materials may need to be used across different lines.

Lean Logistics

The Lean materials computerization must be aware of the physical attributes of the material delivery system, whether this is an automatically guided vehicle (AGV) or a person with a trolley. The timing of journeys for material pickup and put away, the various stages of verification, as well as the capacities of transport, need to be part of the logistics job calculation and assignment. Lean materials computerization will also optimize the traveling route and sequence of replenishment.

In a high-mix environment especially, there can be many peaks and troughs of material logistics requirement. Traditionally, the logistics manpower is calculated to handle peak demand, which creates a lot of waste at other times. With advanced Lean material computerization, the look-ahead capability of the pull signal can be extended to identify potential peaks and to effectively smooth them out by picking materials slightly further in advance of a major changeover.

A reduction of almost 30% of logistics cost can be realized by this method compared to the traditional push system. Picked materials must be delivered to the exact point from which the pull signal originated. Linking the material verification process to the exact reel ID prevents use of material in other locations, even where materials with the same part number are being used.

Lean Warehouse Management

The ability to pinpoint the location of any material at any time is essential to make the material pick process reliable and efficient. Using the unique ID of each material with uniquely labeled warehouse locations, random material storage can easily be supported by the Lean materials computerization. The optimized decision of put-away location, instruction to go to the pick locations, etc., can all be automated so optimization is done for the warehouse location space and the journeys to pick material when pulled. Handheld terminals are ideal for operator guidance.

Inventory accuracy and integrity are also maintained. The computerization knows the position of each storage location in 3D space, as well as its capacity.
Special zones may be set out for hazardous, customer-specific materials, or other reasons such as tax codes. Random location assignment means locations are not restricted to a designated part number, and part numbers can even be mixed in a location. Any specific reel of materials can be easily put away and then later easily found. A continuous cycle check of the warehouse inventory can be performed during times when not all the logistics operators are busy, to eliminate any potential manual errors—that is, materials taken or moved without using the system.

The entire Lean materials process described here, at least for SMT, is a highly specialized process that requires live IoT data from the SMT machines and a fairly complex computerization. Generating the pull signal, optimizing logistics, picking and delivering material, and warehouse enhancement are all an integrated part of the solution. The solution is also complementary to software provided by machine vendors, including, for example, the verification of materials onto feeders and at the machines. This level of integration is not seen in generic MES or ERP solutions.

The use of computerized automation needs sufficient precision that no routine manual intervention is required so data and system integrity is maintained automatically. Any deviations in the operation of the computerization, or lack of ability to correctly model the operation digitally, will cause distrust.
Countermeasures then often lead to system corruption and, in some cases, loss of production. For this reason, incomplete or immature Lean materials systems can cause issues that prevent confidence being attained. While machine communication becomes standardized using IoT technologies, such as the Open Manufacturing Language (OML), and the growth in the demand and expectation of Smart factory solutions increases, the trust in such automation of critical business processes can grow.

Michael Ford is marketing development manager, Mentor Graphics (mentor.com); michael_ford@mentor.com. His column runs bimonthly. For more on OML, visit omlcommunity.com.

Submit to FacebookSubmit to Google PlusSubmit to TwitterSubmit to LinkedInPrint Article