Continuous flow manufacturing.
CFM is a variation of just in time (JIT) manufacturing. CFM was developed by IBM in response to the difficulties it had encountered in implementing JIT. While JIT has been successful in eliminating waste in large production batches, it has been a difficult program to implement.(1) CFM focuses on waste elimination, continuous improvement and company wide involvement. Under CFM, improvements are achieved in months as opposed to years. CFM principles apply to both high and low volume manufacturing lines. CFM focuses on reducing cycle times, continuous improvement and work in-process (WIP) reduction. CFM delineates the relationship between WIP and cycle time. WIP can not be reduced without decreasing cycle time, and vice versa. Of the three, cycle time reduction offers TABULAR DATA OMITTED the greatest leverage and competitive advantage. Cycle time reduction encourages the analysis of work paths, work organization, and the value-added contribution of all process steps, thus fostering change, innovation, and creative thinking in how work is organized and conducted.
Traditional manufacturing pushes the product through the manufacturing process in batch sizes large enough to meet present and future demands. Batch quantities may be determined by a marketing forecast, customer requirements, monthly requirements or any combination of the three. In the pull system, manufacture of the product is linked to customer demand, producing the right product, at the right time, and in the right quantity. The demand in the pull system is met through the use of kanbans and predictable cycle times by controlling the amount of WIP on the manufacturing floor. A kanban is a visual trigger that indicates what to build, when to build and how many to build. Operations are linked together in a chain with kanban buffers. The rate at which each operation must produce is the Takt rate. Takt rate is determined by dividing the daily demand into the number of available work minutes per day. The Takt rate sets the natural rhythm at which the line produces. Some of the key differences between a pull and a push manufacturing process are listed in Table 1.
CFM requires a commitment from everyone involved in the manufacturing process to a new way of doing things. Management must be willing to shut down the manufacturing line when a problem is encountered and to ensure that the necessary resources are available to correct the problem that has shut down the manufacturing line. There must be a penalty free environment for trying new ideas or different approaches. The Lone Ranger no longer rides, and the team or manufacturing cell members are empowered to make decisions that previously were made only by management. There are significant role changes that take place between management and line worker. Management takes more of a coach or mentor role towards the manufacturing cell. The manufacturing line or cell takes ownership for the quality of the work and in meeting the customer's shipping requirements. There is more sharing of roles and expectations with a high emphasis placed on training and meeting the customer's requirement.
Before implementing a pull system, a baseline must be established. A rigorous analysis of the existing manufacturing line or process is undertaken. The existing process is mapped, thereby yielding a detailed process flow. The mapping describes the existing manufacturing process in terms of overall cycle time, WIP, demand, number of process steps, work flow, queue times, nonvalue added work, yields, and other environmental factors. Particular attention is taken in locating bottlenecks or Herbies(2) in the manufacturing process. Bottlenecks determine the rate at which the line can make product. If the demand for a product is 100 units per week and the bottleneck restricts the output to 50 units a week, there is no amount of hard work that is going to deliver the extra 50 units.
In many manufacturing environments, WIP is used to compensate for poor yields and bottlenecks. Under CFM, poor yields and bottlenecks become readily apparent as the WIP level declines. Manufacturing processes with few steps are relatively straightforward in determining where the bottlenecks are and where they will show up as the WIP level is reduced. On more complex manufacturing lines or processes, the initial bottleneck is relatively easy to find. WIP is usually found piling up in front of the equipment or work station. Discovering the subsequent bottleneck as the WIP level continues to decline is not as straightforward. For instance, the initial bottleneck in an assembly process may be at test, but then the next bottleneck needs to be identified. A trial and error approach could be used to reduce the WIP and reveal the next bottleneck, but the risk becomes greater that the line will shut down due to insufficient WIP. WIP in most manufacturing systems masks inefficiencies and yield problems.
In order to determine where the next bottleneck will occur, a detailed line analysis is performed. A software modeling program, such as RapidCycle,(3) accurately models the manufacturing process. The power behind the software is its ability to identify bottlenecks, determine batch sizes, and run a time icon simulation of the manufacturing process. The software real time run mode depicts the manufacturing process and simulates the flow of work throughout the manufacturing process. It is easy to see where the bottlenecks are and how product moves through the process. The models developed from the program accurately predict the process to 90 percent certainty.
One of the measures determined from the process mapping is the manufacturing cycle efficiency (MCE). MCE is a measure of the relative efficiency of a process and is defined as the value add time divided by the total manufacturing time. Value add time is the time spent working on the product where the resulting work changes the form, fit or function. Rework is not considered part of value add and is included in the total manufacturing cycle time. In the overall manufacturing cycle, there are two types of delays that need to be clarified; a lot delay, in which the parts are waiting until the entire batch is completed before they are moved, and synchronizing delay, where parts sit waiting for a piece of equipment because their arrival time was wrong.
While the equation MCE of value added time divided by total manufacturing cycle time, expressed in percent, looks relatively benign, it provides some interesting insight into the overall manufacturing process. MCE for push systems is typically a few percent. When the MCE is calculated for the first time, there is usually concern expressed about the resulting low value. MCE should be viewed as a relative figure of merit that allows various processes to be benchmarked. Many manufacturing improvement programs focus on reducing the value add time and usually result in minimal improvement to the MCE value.
For example, it has been determined that a manufacturing process has an MCE of one percent. The value add time is 10 minutes, and the total cycle time is 1000 minutes. There is a focused effort to reduce the value add time from 10 to five minutes. While this looks like a significant improvement, the MCE declines. The value add time goes from 10 minutes to five minutes. The overall cycle time goes from 1000 minutes to 995 minutes, but the MCE goes from one percent to a half percent.
Many improvement programs focus on improving the efficiency of the worker and ignore the overall manufacturing cycle. Another approach is to focus on queue, setup and transportation times. In many manufacturing processes, the product sits for long periods of time in queues waiting to be worked on. The value added time at the work station may be only a few minutes and then the part is moved on to wait in another queue. Setup times can take anywhere from several minutes to several days. The distance parts travel may add several hundred minutes to the overall cycle time. Take the same manufacturing process with an MCE of one percent and focus on the nonvalue added portion of the overall manufacturing cycle. In this case, the focus is on minimizing queues, and in reducing setup and transportation times. The overall manufacturing cycle time goes from 1000 minutes to 100 minutes, and the MCE improves from one to 10 percent.
Improvement programs that address the overall manufacturing cycle offer the greatest improvement in MCE. When the focus is on reducing queue, setup and transportation times, or other non-value steps, MCE improves approximately 10 to 15 percent. MCE in the 35 percent range is found on processes that are considered world-class. Improvements in MCE are driven by the desire to decrease cycle time.
As part of the mapping process, the manufacturing cell is laid out and the work flow diagrammed. A simplified manufacturing layout that does not include details at each operation and work loop is shown in Figure 1. The path of the part as it moves through the process is carefully reviewed. The distance the part travels, and the number of hand-offs, queues, bottlenecks and shared resources of the process are determined. Information relative to yields and times at each work station are derived by sitting with the cell members. After the data are collected and analyzed, the manufacturing cell is reconfigured. Particular attention is paid to minimizing the distance the parts travel and sit in queues. Set-up times are minimized or eliminated whenever possible so that the parts can immediately be placed into the next operation and have value added to them.
There are some manufacturing processes that do not make sense to incorporate into a work cell. Process steps that require large environment chambers or burn in racks are examples of activities that may not make sense to incorporate into a work cell. These items become part of a shared resource pool used by other manufacturing work cells. Work is scheduled through the shared resource as if it were part of the manufacturing cell.
In addition to mapping the existing process, the skills required at each work station are determined and matched against the available resources. Each work cell member is trained to perform the operation preceding and after his/her work station. Table 2 lists a skill matrix for a typical work cell. Cross training is critical in the overall CFM design. CFM focuses on keeping the work flowing rather than in keeping people busy. Therefore, cross training is a key ingredient in the implementation and evolution of any CFM work cell. As new skills are acquired by the team members, job rotation within the work cell and among other work cells becomes possible. Cross training and job rotation are necessary to create a flexible manufacturing cell focused on keeping work flowing.
One of the first steps in implementing CFM is to lay out the manufacturing floor to minimize the distance that parts need to travel and sit in queues. A typical CFM layout is shown in Figure 2. Stock is moved to the manufacturing floor so that kits are no longer delivered to the work cell. Cell members pull the necessary piece parts from the floor stock and relieve the inventory through bar coding or by making a computer entry.
The amount of WIP on the floor is controlled through the use of kanbans. Kanbans are used to trigger what is to be built, when to build it and how much should be built. There are many different ways to create kanban triggers. A kanban trigger can be as simple as a card or empty bin. A designated area on a bench or floor may serve as a kanban. The kanban area is identified by using tape, coloring the area, having a bin placed in the area, or with some other indicator. The status of the kanban indicates when it is time to build product. An empty kanban signals that it is time to reorder or start work again. Linking the various kanbans together is used to control the quantity of WIP on the manufacturing floor. When a kanban is full, there is no demand for product and work stops. At this point, the resources within the work cell move to other work stations within the cell or use the time for training.
Under CFM, ownership for the work is with the cell members. They are trained to self inspect their work. When the work is moved to the next work station, the operator at this station checks the work. If there should be a problem, the work is immediately returned to the previous work station for corrective action. This procedure provides immediate feedback to the previous work station that there is a problem and corrective action can be taken. With less WIP in the CFM work cell, the assembly processes are moved closer together reducing the distance the parts travel and freeing up floor space. The line is laid out in a U shape to allow the cross trained operators to rotate jobs within the work cell.
Cycle Time and WIP
A number of relationships are used to determine the level of WIP and cycle time interval in the CFM work cell. Cycle time equals WIP divided by exits establishes the basic relationship between WIP and cycle time. A baseline cycle time is determined by performing a WIP over exits calculation. Exits are defined as the quantity of work that leaves the manufacturing cell in a given period of time. Another way exits can be viewed is from the customer's point of view. The customer has a need for product based on a known delivery schedule. Knowing the customer demand (exits) and the manufacturing cycle time determines the amount of WIP necessary to meet the customer's demand.
Decreasing the WIP on the manufacturing floor is an iterative process and is accomplished by improving the overall work flow to match customer demand. From the detailed analysis performed, the initial bottleneck and at what rate this bottleneck produces or passes product down the line becomes known. Resources are assigned to eliminate the bottleneck. Upon eliminating the bottleneck, the next bottleneck will be revealed as the WIP level is further reduced. Again, resources are assigned to eliminate the new bottleneck. This process of WIP reduction and bottleneck elimination is part of the continuous improvement process.
Customer demand in traditional manufacturing is met by pushing product through the process. Run quantities are large enough to satisfy customer needs and are adjusted to compensate for manufacturing inefficiencies. Compensation for manufacturing inefficiencies is achieved by pushing as much work out onto the manufacturing floor as possible. There is a belief that if you cannot see it, you cannot work on it. In some companies where CFM has been implemented, they had to padlock their stockroom door to keep additional WIP from reaching the manufacturing floor. Not seeing piles of work at workstations and having operations idle raises the anxiety level of many managers and supervisors. The anxiety is based upon the belief that shipments will be missed if they do not have WIP piled to the ceiling and beyond. Their focus is on WIP management and not cycle time reduction. Not putting additional WIP out on the manufacturing floor is one of the paradigms that must change and one of the more difficult ones to change. Having less WIP on the floor has another benefit; it frees up working capital. With less WIP on the manufacturing floor, less material needs to be ordered and stored. The amount of freed up working capital can amount to many thousands of dollars.
Importance of Cycle Time
The reasons cycle time is so important must be examined. Cycle time can be thought of in terms of the competition. For example, the manufacturing cycle time to produce a widget is 13 weeks. It is determined that the competition is delivering the same widget and quoting two weeks delivery. In addition, the competition is about to release a series of upgrades to the widget a year before your first upgrade will be available for evaluation. Just as this information has been digested, the sales manager announces that the order for a million widgets has been awarded to the competition because they can deliver the entire order seven months before you even have available capacity to build.
CFM employs small batch sizes and significantly reduced cycle times. Together, cycle time and small batch sizes can be used to increase the rate of learning and improvement that takes place. If a change needs to be made, implementation can be in a matter of hours or days as opposed to weeks or months. If there should be a materials problem, only a handful of units will be built before the problem is discovered. Cycle time can be described in baseball terms. If you get to bat once a season, your ability to improve or hit the ball is greatly diminished. If you get up to bat a few hundred times in the season, your chances of hitting the ball are greatly increased and you may
TABLE III MANUFACTURING CYCLE TIMES CFM Traditional "Pull" "Push" Product Line A Work in Process 300 units 2500 units Manufacturing Cycle Time 4 days 4 weeks Manufacturing Cycle Efficiency 10% [less than] 1% Product Line B Work in Process 2500 units 20,000 units Manufacturing Cycle Time 1 week 16 weeks Manufacturing Cycle Efficiency 15% [less than] 0.08% TABLE IV CFM APPLIED TO BUSINESS PROCESSES Plan & Procure Material Delivery Order Entry Parts Value Add 1 min 9.1 min 105 min Total Time 1754 min 5169 min 11,857 min Cycle Efficiency 0.06% 0.18% 0.89% Major Task 70 170 83 Departments 13 11 5 Transports 14 33 11 Delays 19 35 23 Distance Traveled 2000 ft 9500 ft 3700 ft TABLE V A NINE-MONTH CFM IMPLEMENTATION EFFORT WIP Inventory 53% decline Raw Material Inventory 41% decline Assembly Time 98% decline Shipments 47% increase Yield 183% increase Margin 525% improvement Productivity 44% improvement
even hit a few home runs. Table 3 lists two products manufactured under CFM and under the more traditional push system.
The principles behind CFM can be applied to processes other than manufacturing. Any process that can be mapped can be evaluated in terms of CFM techniques. Three nonmanufacturing processes are listed in Table 4. The MCE for all three processes was less than one percent before the principles behind CFM were applied. In the case of plan and procure parts, the paper work traveled 9500 feet and took over two weeks to complete its journey. There were 170 major tasks required along the way, the paper crossed 11 different departments for approvals, was transported 33 times, and sat in various in boxes 35 times. The value add work that actually took place during the two week cycle amounted to less than 10 minutes.
After applying the principles of CFM to the plan and procure process, the MCE improved from 0.18 percent to 15 percent. Under the continuous improvement process, the MCE for planning and procure parts has improved from 15 percent to 25 percent.
CFM can be implemented in a relatively short period of time and does not require a significant capital investment. Under CFM, a flow or rhythm is established that is adjusted to the customer's demand. CFM does require a change in attitude towards the manufacturing process. The benefits of less WIP and reduced manufacturing cycle time begin immediately upon implementing CFM. The most difficult part is to adhere to the disciplines, and if a commitment is made to CFM, it is a commitment for continuous improvement and change. Table 5 lists a nine month CFM implementation effort. This operation manufactures discrete RF power transistors and modules used in the communications and radar markets. CFM has been implemented in other facilities on a worldwide basis. The same kinds of improvements have been experienced after implementing CFM. As the methodology is refined, further improvements have been possible in cycle time reduction.
The author gratefully acknowledges the inputs provided by Bob Lynch of M/A-COM and Chris Owens of George Group Inc. Much of the work and initiatives surrounding CFM at M/A-COM would not have been possible without their involvement and commitment. There have been countless other individuals that have made CFM part of the culture at M/A-COM.
1. James Gooch, Michael George and Douglas Montgomery, "America Can Compete," George Group, 1987, p. 6
2. E.M. Goldratt and J. Cox, The Goal, North River Press, 1986, pp. 112-122.
3. RapidCycle is a registered trademark of George Group Inc.
Jack Hillson received his undergraduate degree in physics from Florida Institute of Technology. He received his masters of business administration from Rivier College. Hillson has worked in a number of M/A-COM operations during the last 10 years, and is currently the director of manufacturing support for M/A-COM's Microelectronics Division. His current interests involve flexible manufacturing and business process reengineering. Hillson is an IEEE member.
|Printer friendly Cite/link Email Feedback|
|Title Annotation:||Special Report; just-in-time manufacturing|
|Date:||Oct 1, 1994|
|Previous Article:||Practical high temperature superconductor filters.|
|Next Article:||Unwarranted investigations: rules to live by.|