Connecting Assembly to Batch Processes through Basic Pull Systems

Portuguese translation by Diogo Kosaka.

Pull systems are a fundamental part of lean manufacturing and are frequently misunderstood and considered difficult to implement. Specifically, one frequent area of difficulty, I believe, is finding the proper connection between assembly line processes and batch processes such as stamping, injection molding, painting, or operations such as machining, for example.

There are three basic types of pull production systems: the supermarket pull system, the sequential pull system, and the mixed pull system, which combines elements of the other two (see the glossary at the end of the story). In all three cases, the technical elements that are important for the system to succeed are:

1. Products flowing in small batches (creating “one piece flow” wherever possible).

2. Pulling processes according to takt time (not allowing overproduction).

3. Signaling replenishment through a kanban.

4. Leveling the product mix and quantity.

To connect batch processes with the assembly line, the pull production system is used together with a specific type of kanban known as a signal kanban. There are at least three different types of signal kanban, and the appropriate use depends on the nature of your process and manufacturing. The three types of signal kanban are:

1. Standard production kanban

2. Batch-making kanban

3. Triangle kanban

The Standard Production kanban establishes a fixed production order and some basic rules on batch size, which may vary somewhat with need. However, the fundamental pattern, or production order, is always maintained. This creates a fixed (i.e., standard) production sequence that is continuously repeated. This style of production is generally necessary in heat treatment, painting, or other processes with specific changeover sequences. While the product sequence should be fixed in these cases, the actual quantity produced in each cycle is not fixed and varies according to customer needs. This is known as a fixed order/variable quantity replenishment cycle.

For example, in an eight-hour cycle, parts always go from A to F (the difficulty of the changeovers may dictate this order). Central supermarket inventory is a function of the replenishment cycle pattern. Thus, a one-day cycle implies a minimum of one day of inventory that must be kept in the supermarket; a one-week cycle means one week of inventory. The main disadvantage of the Standard Production Kanban is that the sequence is fixed; you cannot jump from part C to part F.

The Standard Production Kanban establishes a production rhythm and sequence. As long as the pace of all production is kept close to takt time and the supermarket drives the pull between the two processes, the system works. It is the first and basic step toward establishing production control and a pull system between two processes separated by a supermarket.

The pros are that it creates an established order where none existed before. It also allows the best production sequence to facilitate changeovers, if necessary. The cons are that it is not a perfect just-in-time system, since you must wait your turn in the order as the queue moves, so some delay is involved. Unfortunately, it also does not tend to pressure the system to reduce changeover times, inventory time, or lead time (except through management intervention).

Batch control board

Making batches with a batch control board requires creating a physical kanban for each container of parts in the system. As soon as material is consumed from the supermarket, the kanban is immediately detached and periodically brought back to the upstream batch production process on a board that lays out all the parts and shows the empty space between each kanban card in the system. This is an example of the fixed quantity/variable order style of pull replenishment.

The return of the kanban to the empty space on the board indicates that inventory has been consumed from the supermarket; kanbans not returned represent inventory still in the supermarket. Once the predefined trigger point is reached, the production operator knows to start making the product to replenish the material in the supermarket. If several trigger points are reached at once, production has to establish rules regarding what should be produced first (in this scenario the pattern or sequence is not fixed) and where to produce it. Otherwise, supervisors will decide what to do based on the information available.

There are several advantages to this method. One is that the scheduling sequence is more flexible than with the Standard Production Kanban. A second is that it provides a visual representation of inventory consumption and highlights emerging problems in the central supermarket.

There are disadvantages to this type of kanban as well. It requires many kanbans if you have many part types, and the kanbans must be brought back in a timely and reliable manner for the batch control board to be accurate. In addition, there is the inherent problem of several parts reaching the trigger point at the same time. In these circumstances, human judgment is needed to decide what to produce among two or more products that have reached the trigger point. Discipline is required on the part of schedulers and supervisors not to allow inventory to be built ahead of when it is needed.

Triangle Kanban

The method Toyota would follow most of the time to link batch processes with the downstream assembly process is called the triangle kanban. Below is an example, without much detail, of how to use this method in a lean implementation (see the LIB workbook Creating Level Pull for more detailed explanations). The example assumes you have a batch machining department that feeds a final assembly line. The machining department has five machines that make 11 completely different product variations. Demand for each product varies significantly, as does the standard packaging for each part. Setup (changeover) time is roughly three hours.

1. Take the 11 parts and dedicate them to the five machines. (Ideally, for example, four machines would have two part types to make and one machine would have three part types to make, but this depends on volume and mix, etc.)

2. Now let us take one dedicated machine that has several parts assigned to it. 

3. Create a single triangle kanban for each part, including the following information.

a. Part number

b. Part description

c. Inventory location

d. Machine on which it is to be made

e. Space to record the trigger date

f. Tool number

g. Batch size

h. Trigger point

4. Determine the daily quantity to be made of each part. Calculate your average daily demand for each part. If you have not leveled the final assembly line, be prepared to add an extra quantity to cover variation in demand. Let us assume you have two eight-hour shifts available for production, equivalent to 16 hours of production (the simplest case). If you have two parts on one machine, you can size the production time based on average daily demand and the cycle time to make the part. For example, let us say, to make the calculations easy, that the required production time is 10 hours.

5. Determine the time available for changeovers. In this case, 16 hours of production and 10 hours of demand leave six hours per day for changeovers. You must factor in any downtime and scrap in the short term – but eliminate them in the long term. So, in our case (ignoring “downtime”, etc.) you have time available for two changeovers per day, or roughly one per shift according to current calculations.

6. Determine the batch size for each part. There are several different ways to do this, but for simplicity I will describe only the easiest. In this case, your batch size is just one day of production, since you have two part types for this machine and will average two changeovers per day. In short, you are making every part every day. If you had 10 part types on this machine, you would have batch sizes of five days.

7. Determine your replenishment trigger point. To do this, add the time to make the longest-running component, add the changeover times, and add the time to make the first container and put it back in the supermarket. This is the minimum level you can set as a trigger point for replenishing a product. (An average trigger point might be 300 units.)

The system works by hanging a single triangle kanban at the trigger point in inventory. When the trigger point is reached, the material handler takes the kanban back to the machine for production. The kanban is hung on the machine and dictates what is to be produced and the batch size.

The triangle pull system has many advantages. It runs on its own once established, as long as average demand does not change. If it does change, change your batch size as needed. The triangle system also helps establish quality, since the product is dedicated to one machine, and it absorbs some variation, but you can still choose to make the product on other machines.

The downside is that supermarket inventory is not visible at all times, since there is no batch control board with cards. You will see, however, how many kanbans are hanging on the machine, which is a good indicator that inventory is running low and also a tool for visual control. In addition, there is only one kanban per part to control. Other minor challenges include, of course, understanding when the deadline for scheduling maintenance should be (for some machine types) and what impact this will have on scheduling. You should also watch for a simple signal to bring any necessary raw material to the machine.

In the end, you must decide what your main objective is in implementing the pull system. Each of the three methods described works for scheduling batch processes together with the supermarket, but each emphasizes a main point. It is worth calculating inventory levels, cycle times, and operating rules for each of the three different methods. Build a simple matrix, evaluate each of them according to your priorities, and choose the one that best fits your company’s needs.

Glossary (adapted from the Lean Lexicon)

There are three basic types of pull system:

Supermarket Pull System
This is the most basic and widespread form of pull production, also known as a replenishment pull system or type A pull system. In a supermarket pull system, each process has a store, a supermarket, that holds a quantity of each item produced. Each process produces only what is needed to replenish what is withdrawn from its supermarket. Typically, when material is withdrawn from the supermarket by the downstream process, a kanban or other type of information is sent upstream to the supplying process, which will then replenish what was withdrawn. Each process is responsible for replenishing its own supermarket, so daily management of the workplace is relatively simple and kaizen opportunities are relatively easy to see. The disadvantage of a supermarket system is that a process must hold an inventory of every part it produces, which may not be practical if the variety of parts is very large.

Sequential Pull System
Also known as a type B pull system, the sequential pull system can be used when there is too wide a variety of parts to be stored in a supermarket. Products are essentially made to order, and overall system inventory is minimized.
In a sequential system, the scheduling department must set the correct mix and quantity of products to be produced. This can be done by placing kanban cards in a Heijunka Box, usually at the start of each shift. These production instructions are then sent to the first process in the value stream. It is common for this to be done in the form of a “sequential list”. Each of the following processes produces, in sequence, the items that arrive from the previous process.
FIFO for each product must be maintained continuously.
A sequential system creates pressure to keep lead times short and predictable. For this system to work effectively, the pattern of customer orders must be well understood.
If orders are difficult to predict, the production lead time must be very short (shorter than the order lead time) or an adequate supermarket of finished goods must be maintained.
A sequential system requires strict management to sustain it, and improvements on the shop floor can be a difficult challenge.
 
Mixed Supermarket and Sequential Pull System
Supermarket and sequential pull systems can be used together in a mixed system, also known as a type C pull system. A mixed system may be appropriate when an 80/20 rule applies: a small percentage of the range of parts (perhaps 20%) accounts for most (perhaps 80%) of daily production volume.
Typically, an analysis is done to segment parts by volume, according to order frequency: (A) high, (B) medium, (C) low, and (D) infrequent. Type D may represent orders or special service parts. To handle slow-moving items, a special kanban for type D can be created to represent not a specific part but a certain amount of capacity. The production sequence for D products is then set by the method used by the scheduling department for parts in the sequential pull system.
Such a mixed system allows supermarket and sequential systems to be applied selectively, gaining the benefits of each, even in environments where demand is complex and varied. The two systems can operate at the same time, side by side, horizontally, along an entire value stream, or they can be used for a given part in some segments of its specific value stream.
A mixed system can make it harder to balance work and identify abnormal conditions. It can also make conducting and managing kaizen more difficult. Therefore, for the mixed system to work effectively, very rigorous management is required.