Slashing lead-times in Belgium

https://commons.wikimedia.org/wiki/File:Time-Allocation.png
ISE Magazine November 2018 Volume: 50 Number: 11
By Pascal Pollet and Ben Proesmans

A few years ago, Provan, a small metalworking subcontractor in Genk, Belgium, found that it had become a victim of its own success. Provan manufactured stoves for a major customer in Europe. The customer was pleased and wanted Provan to increase the variety of stoves it made. When reviewing the requirements, Provan’s management realized that with its current manufacturing methods, this increased variety would require tripling warehouse space for intermediate components. The added complexity of production operations would also lengthen lead-times, increasing the work in process (WIP) and storage space for semi-finished parts.

All this additional space would come at a significant cost in Belgium. And most critically, the escalating investment in this inventory would burden Provan’s cash flow – potentially crippling a small enterprise.

Around this time, managers started learning about quick response manufacturing (QRM) and paired-cell overlapping loops of cards with authorization (POLCA). Managers thought these approaches, pioneered by professor Rajan Suri at the University of Wisconsin-Madison, might help increase the variety of stoves made by Provan without requiring substantial investments. This is the story of how applying QRM and POLCA worked.

Provan’s previous production system

After its founding in 1998 by co-author Ben Proesmans and Luc Vanhees, Provan emerged as a successful metalworking subcontractor and supplier of metal products. The company now consists of around 80 employees and serves a variety of industries, including medical, automotive, agriculture, heating and machinery. Provan offers its customers a total solution for welded structures, laser and sheet-metal work, profile machining and assembly.

As the company grew, management noticed that Provan’s customers were demanding a lot more variety, and simultaneously, batch sizes in the orders had shrunk considerably. Yet at the same time, customers wanted more flexibility and shorter delivery times. Provan had been using its enterprise resource planning (ERP) system together with lean techniques such as kanban to manage the order flow. However, the analysis by Provan’s management showed that an increase in variety would make these material management systems unsuitable for Provan’s production characteristics.

The issues with the existing systems can be illustrated by the example of producing stoves for the customer mentioned above. The stoves consisted of 130 parts that were produced in several steps such as laser cutting, bending and threading. Then these 130 parts were welded or fastened together and assembled into one stove. The parts were stored between the successive fabrication and assembly steps in a warehouse that occupied more than 600 square meters (around 6,500 square feet). The stoves were produced in batches of 60, and the lead-time from the initial laser cutting operation until the final inspection took about four weeks.

As mentioned, the customer was pleased with Provan’s performance and asked the company if it could supply three variations of the stove product. This meant that three types of stoves needed to be welded and assembled, and 130 components per stove – often with different routings – needed to be manufactured. As mentioned earlier, management calculated that this would triple the needed warehouse space to store the components, along with adding the other problematic issues.

Implementing a QRM cell for the stoves

Management thought that Suri’s QRM approach could help resolve production problems even in the presence of high variety, and the company decided to create a QRM cell for the production of stoves. QRM cells are based on the well-known concept of cellular manufacturing but incorporate some additional features specific to QRM. (See the sidebar below.)

In creating a QRM cell for the stoves, it was not clear how to manage the complexity of the workflow within the cell. Since three types of stoves needed to be fabricated and 130 components had to be produced for each stove – typically with different routings within the cell – there was concern about how the operators could manage all these material flows. Would Provan have to install a miniature version of an ERP system within the cell? That move would bring additional complexity and other scheduling problems related to the functioning of such an ERP system.

After a few months of brainstorming, Provan’s employees arrived at a solution that would be simple to use and might avoid the typical problems associated with complex ERP systems. Serendipitously, it turned out that an additional benefit was that this solution enabled the employees to become familiar with visual management and using color-coding to signal tasks and priorities, preparing them for the companywide POLCA implementation that took place later (and is described below).

The idea involved creating a QRM cell that would include all the processing steps after the initial laser cutting. The laser was kept out of the cell because this expensive machine was also used for many products for other clients. Buying an additional laser for the cell couldn’t be justified, and it didn’t seem necessary to accomplish the goals for the stove products. All the remaining workstations – involving operations such as bending, rolling, threading, grinding and welding – were moved into a U-shaped cell.

Visual workflow management through color coding

The next concept in the Provan team’s idea was to assign every workstation a color and then make sure that this color code was clearly marked on each workstation.

This is how the color-coded system works. When parts for a batch of stoves arrive from the laser station, they are put on a set of carts in a standardized way. First, a numbered metal flag with a specific color is attached to an empty cart. The color on the numbered flag corresponds to the stove type. Hence, with three types of stoves, three different colors are used. The number on the flag indicates which parts should be put on that cart. Multiple parts are grouped together on the cart based on their routing within the cell.

Then on every cart, a row of additional colored flags is added. These flags indicate the routing the cart has to follow within the cell. The first color in the flag row corresponds with the color of the first workstation in the routing. The second color corresponds with the color of the second workstation and so on. To summarize, two types of colored flags are used to guide the material flow: The colored numbered flag indicates the product type, while the colors on the (unnumbered) flags specify the routing.

After a cart has been coded with its appropriate set of flags, it is placed in the middle of the stove cell. Based on the first (unnumbered) colored flag on the cart as well as workstation availability, an operator pulls the cart into the appropriate workstation and processes the parts on the cart. After all the parts on the cart are processed at this workstation, the corresponding colored workstation flag is removed by the operator and the cart is put back into the middle of the cell.

The cart is now available for processing at its next workstation and so on. With this system, operators can easily spot which parts they have to process without consulting a computer system or printed shop orders. They just have to look at the colored flags to know which carts are available to be processed next at any given workstation.

This production control system offers many benefits:

• In keeping with the latest manufacturing practices, the work in process (WIP) in the cell is strictly limited by the finite and small number of available carts. Limiting the WIP has several advantages for the cell. Material piling up can lead to all kinds of waste (searching for material, larger walking distances, damage and so on), and so all these wastes are avoided. Also, limiting the WIP in turn limits the lead-time of jobs in the cell.
• The highly visual system is easy for the operators to understand and use.
• The order status information (the status indicated by the flags) is always up to date. So the production control system doesn’t rely on outdated information like it did in the past.
• The workflow is completely self-steering. Employees are in control of the process, which means they feel more involved. Detailed planning of the separate workstations by a planner, as well as supervisory tasks like shifting workers between the work stations, have become unnecessary.

These benefits quickly manifested themselves in terms of significant results for Provan:

• Putting all the post-laser operations in a single production cell meant that material movement on the shop floor was reduced, and stoves could be made in batches of 15.
• The packing process could be done in half the time because the parts were already on the right pallet and weren’t stored in the warehouse waiting to be picked.
• The total lead-time per batch decreased by an amazing 85 percent, from around four weeks to just three days.
• And finally, an unanticipated – but very welcome – result of implementing the cell has been that the quality of the products has improved significantly: Scrap and complaints have been reduced by 60 percent.

The short lead-time and near-perfect delivery performance resulting from implementing the stove cell resulted in more products being ordered by the customer: Provan now makes eight models of stoves. Even with many more models, the on-time delivery performance is still 100 percent.

And one of the best results for Provan’s management has been that instead of tripling the warehouse space (or more), the stocks of parts have actually been completely eliminated. This has liberated 600 square meters of space, allowing Provan to expand its production operations without having to rent more space. Plus, the reduction in stocks freed up much-needed working capital to help finance Provan’s growth.

For the rest, POLCA keeps the job shop in tune

Based on the resounding success of the stove cell, Provan looked for additional opportunities to form QRM cells. Another successful cell that was implemented was an assembly cell for medical treatment couches. Then a third cell dedicated to prototyping work was created.

These three cells represented about 50 percent of the workload on the shop floor. For the remaining 50 percent of the workload, the company operated like a typical job shop. Since Provan served as a subcontractor for a wide variety of customers, the volume and work content of jobs varied frequently, making it difficult to dedicate work to particular cells focused on clear product families.

For these remaining high-variety jobs, Provan suffered from several problems on its shop floor: Information at the work centers was often already outdated, even by the time the production orders were distributed to the shop floor; planning and re-planning activities were a heavy burden; supervisors had to run around the shop floor to expedite orders; and priorities were shifting continuously. Finally, despite all the follow-up and rescheduling, the necessary parts to start an order were often missing. As a result, the operators did not trust the order information.

Provan’s approach to solve these issues for the remaining 50 percent of the work consisted of four steps:

• Even though it was not possible to create cells for entire products (as was done for the stoves), Provan was able to rearrange the shop floor into manufacturing cells in which similar operations were grouped together.
• The high variety of customer products was then accommodated by routing jobs through different combinations of cells based on their needs.
• The planning was simplified. Previously, planning and order routing was done at the detailed level of individual machines. Now, the planning is done at the higher level of cells, and operators in the cells are given ownership and responsibility within the cells so that they can handle the work themselves in a more flexible way.
• Finally, the workflow between the cells was controlled by using the POLCA system.

POLCA, remember, stands for paired-cell overlapping loops of cards with authorization. This card-based visual control system manages the flow of jobs through the shop floor. At each of the cells it controls which job should be worked on next in order to meet delivery targets. POLCA ensures that upstream operations use their capacity effectively by working on jobs that are needed downstream, while at the same time preventing excessive WIP build-ups when bottlenecks appear unexpectedly. POLCA is particularly effective in low-volume, high-mix and custom production environments.